Polycomb repressive complex 2 is required for MLL-AF9 leukemia

Tobias Neffa,b, Amit U. Sinhaa,b, Michael J. Klukb,c, Nan Zhua,b, Mohamed H. Khattaba,b, Lauren Steina,b, Huafeng Xiea,b, Stuart H. Orkina,b,d,e,1, and Scott A. Armstronga,b,e,1

aDana-Farber/Children’s Hospital Center, Boston, MA 02115; bHarvard Medical School, Boston, MA 02115; cDepartment of Pathology, Brigham and Women’s Hospital, Boston, MA 02115; dHoward Hughes Medical Institute, Boston, MA 02115; and eHarvard Institute, Cambridge, MA 02138

Contributed by Stuart H. Orkin, February 8, 2012 (sent for review January 5, 2012) A growing body of data suggests the importance of epigenetic plagued by off-target effects, especially when the readout is mechanisms in cancer. Polycomb repressive complex 2 (PRC2) has cellular toxicity. Here we describe studies evaluating loss of been implicated in self-renewal and cancer progression, and its com- PRC2 function using conditional alleles, allowing for specific and ponents are overexpressed in many . However, its role in complete deletion of individual PRC2 components. We con- cancer development and progression remains unclear. We used con- ducted these studies in a well-defined mouse model of leukemia ditional alleles for the PRC2 components enhancer of zeste 2 (Ezh2) driven by the MLL-AF9 fusion oncoprotein. and embryonic ectoderm development (Eed) to characterize the We show that genetic inactivation of Ezh2 compromises but role of PRC2 function in leukemia development and progression. does not completely abrogate leukemia growth. Ezh2-inactivated Compared with wild-type leukemia, Ezh2-null MLL-AF9–mediated MLL-AF9 leukemia shows persistence of on many acute myeloid leukemia (AML) failed to accelerate upon secondary loci, suggesting compensation by Ezh1, as previously described in transplantation. However, Ezh2-null leukemias maintained self-re- embryonic stem (ES) cells (16). In contrast, complete genetic newal up to the third round of transplantation, indicating that loss of PRC2-function by inactivation of Eed is incompatible with Ezh2 is not strictly required for MLL-AF9 AML, but plays a role in leukemic self-renewal. Our results demonstrate an absolute re- leukemia progression. Genome-wide analyses of PRC2-mediated tri- quirement for PRC2 function in MLL-AF9 AML and define methylation of 3 demonstrated locus-specific persistence of a role for Ezh2 in cancer progression. H3K27me3 despite inactivation of Ezh2, suggesting partial compen- MEDICAL SCIENCES sation by Ezh1. In contrast, inactivation of the essential PRC2 gene, Results Eed, led to complete ablation of PRC2 function, which was incom- Inactivation of Ezh2 Interferes with Growth of Preleukemic Colonies patible with leukemia growth. Gene expression array analyses ex Vivo. We characterized the effects of genetic inactivation of indicated more profound gene expression changes in Eed-null com- PRC2 components in a defined model of AML. To this end, we pared with Ezh2-null leukemic cells, including down-regulation of used mice carrying three genetic modifications: (i) homozygously Myc target genes and up-regulation of PRC2 targets. Manipulating floxed sequences for either Ezh2 or Eed,(ii) polyI:polyC (pIpC)- PRC2 function may be of therapeutic benefit in AML. inducible Cre (MxCre), and (iii) a Cre reporter, flox-STOP-flox- ROSA26-YFP (17) (Fig. S1A). Initially, we studied the effects of − | mouse model | polycomb group proteins | loss of Ezh2 function in vitro. Lineage marker negative (Lin ), myeloid-lymphoid leukemia protein c-Kit+, and Sca-1+ (LSK) immature progenitors and stem cells were purified by flow sorting, and cells were transduced with olycomb repressive complex 2 (PRC2) has been implicated a bicistronic γ-retroviral vector encoding MLL-AF9 and linked Pin development and cancer (1, 2). PRC2 is composed of the via an internal ribosomal entry site (IRES), the reporter, GFP. core components embryonic ectoderm development (EED), sup- Cells were transduced with a self-excising retroviral vector pressor of zeste 12 (SUZ12), and a SET-domain methyltransfer- encoding Cre recombinase (18). Cells double positive for GFP ase, either enhancer of zeste 2 (EZH2) or enhancer of zeste 1 and YFP were sorted 6 d after Cre transduction and plated in (EZH1) (3). The PRC2 complex catalyzes the di- and trimethy- methylcellulose. Ezh2-null cells showed a reduction in colony lation of lysine residue 27 of histone 3 (H3K27me3), a repressive growth (Fig. S1B), with smaller and scattered colonies (Fig. mark (4). Overexpression of EZH2 has been correlated S1C). Cytospins of primary Ezh2-null colonies showed more with prostate cancer progression (5). Follow-up studies have con- differentiated morphology compared with WT counterparts (Fig. firmed and expanded these results for other cancer types, mainly S1C). Attempts to replate Ezh2-null MLL-AF9 cells led to solid tumors. In the hematopoietic system, forced expression of progressive outgrowth of cells with incompletely deleted Ezh2 Ezh2 increases serial transplantation potential in hematopoietic sequences (as measured by qPCR, Fig. S1B), suggesting a com- stem cells (6) and enhances transformation in a model of multiple petitive disadvantage of Ezh2-null cells in vitro. These data myeloma (7). Intriguingly, heterozygous loss of function of EZH2 demonstrate that inactivation of Ezh2 impairs leukemic colony has been associated with adverse prognosis in myelofibrosis (8), formation ex vivo. and heterozygous and homozygous inactivation of EZH2 has been described in myelodysplastic syndromes and more recently in T- lineage lymphoblastic leukemia (9–12). In contrast, EZH2 alter- Author contributions: T.N., S.H.O., and S.A.A. designed research; T.N., A.U.S., M.J.K., N.Z., ations appear to be rare events in acute myeloid leukemia (AML). M.H.K., L.S., and H.X. performed research; A.U.S. and S.H.O. contributed new reagents/ fi analytic tools; T.N., A.U.S., M.J.K., S.H.O., and S.A.A. analyzed data; and T.N., S.H.O., and S. How to reconcile these ndings on a mechanistic level is unclear. A.A. wrote the paper. Given the heterogeneity of clinical samples, elucidating the com- The authors declare no conflict of interest. plex role of PRC2 biology in cancer will be aided by studies in Freely available online through the PNAS open access option. genetically defined animal models (13). Data deposition: The data reported in this paper have been deposited in the Gene Ex- Loss of function of PRC2 has been evaluated in several cancer pression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. GSE34963). models (14, 15). However, shRNA technologies used in these 1To whom correspondence may be addressed. E-mail: [email protected] or studies will leave in place some residual expression and function [email protected]. of the protein in question, obscuring the interpretation of ex- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. perimental results. Furthermore, studies using shRNA can be 1073/pnas.1202258109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1202258109 PNAS Early Edition | 1of6 Downloaded by guest on September 26, 2021 Ezh2-Null MLL-AF9 Leukemia Is Compromised but Maintains Self- leukemia, there was a significant survival advantage in the Ezh2- Renewal Ability. We proceeded to study the in vivo effect of null group (Fig. 1C). Resulting leukemias were typically entirely Ezh2 inactivation on MLL-AF9 leukemia. Blast colonies repla- composed of GFP+/YFP+ double positive cells. Analysis of ted one to two times were derived from MLL-AF9–transduced sorted GFP+/YFP+ double positive cells by Western blot (Fig. flox/flox LSK cells from Ezh2 and WT mice. A total of 1 × 105 cells S2D) and three-primer PCR (Fig. S2E) demonstrated outgrowth per recipient were i.v. injected into sublethally (600 cGy) irra- of incompletely inactivated cells in one of four tested secondary flox/flox diated primary recipient mice. Ezh2 and WT donor mice recipients. We then performed tertiary transplantation of WT also carried the MxCre and ROSA-YFP Cre-reporter alleles. and Ezh2-null leukemia. As in secondary recipients, there was pIpC was administered beginning 2–3 wk after transplantation. a significant survival advantage in the Ezh2-null group (Fig. Two separate cohorts using MLL-AF9–transduced cells from S3A). Q-PCR (Fig. S3B), Western blot (Fig. S3C), and three- two sets of donor mice were studied. Somewhat unexpectedly, primer PCR (Fig. S3D) demonstrated outgrowth of incompletely given the impaired growth of Ezh2-null colonies in vitro, there Ezh2-inactivated cells in a proportion, but not all of the tested was no apparent effect of Ezh2 inactivation on survival in pri- tertiary recipients. These data are consistent with a competitive mary recipients (Fig. 1A). Leukemic mice typically had re- survival disadvantage for Ezh2-null leukemic cells. However, the placement of the bone marrow by leukemic cells of >90%, and results also demonstrate that Ezh2 is not absolutely required for spleens were consistently enlarged. Flow cytometry demon- leukemia maintenance in the MLL-AF9 model, at least not until strated that the majority of GFP+ leukemia cells had excised the the third round of serial transplantation. The phenotype of Ezh2- Cre reporter and had therefore also acquired YFP expression null tertiary leukemia was compatible with AML as determined (Fig. S2A). Western blot of sorted Ezh2-null GFP+/YFP+ leu- by flow cytometry (Fig. S3E). Cell cycle analysis showed a modest kemic cells confirmed absence of Ezh2 protein and demon- but significant decrease in leukemic cells in S phase in Ezh2-null strated substantially reduced, but detectable levels of H3K27me3 vs. WT leukemias in spleen (Fig. 1D) and bone marrow (Fig. S4). (Fig. 1B). Near complete deletion of Ezh2 in leukemic cells from Given the cell cycle alterations in Ezh2-null cells, we evaluated primary recipients was documented by qPCR (Fig. S2B) and by the expression of known PRC2 targets, p16ink4a and p19arf, three-primer PCR (Fig. S2C). Sorted GFP+/YFP+ leukemic encoded by the Cdkn2a locus, by Western blot and found both cells were transplanted into secondary recipients and three doses gene products up-regulated in Ezh2-null leukemias (Fig. 1 E of pIpC were administered beginning 1 wk after transplantation and F). p16ink4a and p19arf are known tumor suppressors with to ensure full deletion of floxed sequences. In contrast to primary roles in the Rb and p53 pathways, respectively, and it is likely that their up-regulation contributes to the phenotype observed after inactivation of Ezh2.

Expression Profiling of Ezh2-Null Leukemia Reveals a Functional Link Between Gene Expression Programs Controlled by PRC2 and Myc. To better characterize the effects of Ezh2 deletion in MLL-AF9 leukemia on polycomb target genes, we performed gene ex- pression profiling of primary and secondary WT and Ezh2-null leukemias. MLL-AF9 leukemia has been linked to an ES-cell– like gene expression signature (19). More recently, the ES-cell– like signature often observed in cancer has been further sub- divided into a PRC2 module, a Myc module, and a core ES module. These modules are defined by combinatorial DNA binding of Eed, Myc, and Nanog, and their associated factors (20). Because we inactivated Ezh2, a PRC2 component, we tested enrichment of the PRC2 module by gene set enrichment analysis (GSEA) and found this signature to be significantly enriched in secondary Ezh2-null AML compared with secondary WT AML (Fig. 2A). Importantly, we found the Myc module genes to be significantly negatively enriched in Ezh2-null leuke- mic cells compared with WT counterparts (Fig. 2B). WT and Ezh2-null tertiary leukemias showed similar degrees of in- filtration of liver and spleen (Fig. S5) but WT tertiary leukemias showed a more prominent infiltration of the kidney (Fig. 2C). In contrast to the PRC2 and Myc modules, no significant negative or positive enrichment of the ES-cell core module signature was observed in Ezh2-null leukemic cells (Fig. S6A). This finding is consistent with prior observations that the ES-cell core module may be of less importance in cancers. Inactivation of Ezh2 also Fig. 1. Inactivation of Ezh2 in vivo is compatible with MLL-AF9 leukemia did not result in enrichment of primary MLL-AF9 binding tar- growth but leads to a less aggressive phenotype. (A) No survival difference gets, as is observed, e.g., after inactivation of the histone meth- between recipients of MLL-AF9–transduced cells with WT or inactivated Ezh2 yltransferase, Dot1l (21) (Fig. S6B). The gene sets used are listed locus. (B) Western blot analysis demonstrates loss of Ezh2 protein in Ezh2- in Dataset S1. null leukemic cells in vivo and substantially reduced H3K27me3. (C) Signifi- Decreased Myc expression has been demonstrated after loss of cantly prolonged survival of secondary recipients in the Ezh2-null group. (D) function of another epigenetic regulator implied in MLL-AF9 BrdU in vivo cell cycle analysis of leukemic WT and Ezh2-null leukemic spleen leukemia, namely, Brd4 (22–24). However, despite significant cells showing moderate cell cycle alterations in the Ezh2-null group (de- Ezh2 crease in S phase). n = 2 (WT) and n =4(Ezh2-null). Error bars indicate range. attenuation of the Myc module, inactivation of in our fi system did not lead to altered Myc protein levels by Western blot G0/G1: P = 0.0007, S: P = 0.0028. G2/M: not signi cant (analysis of bone D Ezh2 marrow and spleen cells, two-way ANOVA). (E) In vivo up-regulation of (Fig. 2 ), or mRNA levels by qPCR (relative expression / p16ink4a in Ezh2-null leukemia by Western blot. (F) In vivo up-regulation of WT = 1). These data suggest an indirect mechanism for the p19arf in Ezh2-null leukemia. observed down-regulation of Myc module genes. It is noteworthy

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1202258109 Neff et al. Downloaded by guest on September 26, 2021 Fig. 2. Analysis of Ezh2-null leukemia. (A) Gene set enrichment analysis

Fig. 3. ChIP-Seq analysis of primary and secondary WT and Ezh2-null leu- MEDICAL SCIENCES reveals enrichment of PRC2 module in Ezh2-null secondary leukemia. (B) kemic cells distinguishes class 1 genes that lose H3K27me3 and class 2 genes Highly significant negative enrichment of Myc module genes in Ezh2-null that retain H3K27me3. (A) Representative plot showing loss of H3K27me3 at compared with WT secondary leukemia. NES, normalized enrichment score. the Bmi1 locus (as an example of a class 1 gene) and retention of H3K27me3 (C) Increased infiltration of the kidney in tertiary recipients of WT leukemias at the Hes1 locus (as an example of a class 2 gene). (B) Genes in class 1 compared with recipients of Ezh2-null leukemic cells. (D) Myc protein levels showed a significant increase in expression in Ezh2-null cells compared with are comparable between tertiary WT and Ezh2-null leukemia. WT cells. In contrast, there is no significant pattern of expression change in class 2 genes. (C) Genes retaining significant H3K27me3 level (P =0.005)after inactivation of Ezh2 overlap significantly with documented EZH1 targets that PRC2 components have recently been shown to be direct (genes bound by EZH1 and genes that retain H3K27me3 in Ezh2-null ES cells). downstream targets of Myc (25). Our genome-wide expression profiling data suggest that the expression of Myc target genes is partly dependent on PRC2, a functional link between PRC2 and whereas inactivation of Ezh2 leads to substantial but incomplete Myc module gene expression that had not been previously noted. loss of H3K27me3 and a more moderate disturbance in pheno- type, owing to partial compensation by Ezh1 (16). To further Progression of WT MLL-AF9 Correlates with Down-Regulation of PRC2 characterize the effects of Ezh2 inactivation in AML, we ana- Module Genes and Depends on Ezh2. Analysis of survival curves lyzed the H3K27me3 pattern for WT and Ezh2-null leukemic shows a progression from primary WT leukemia to secondary cells. To this end, we performed chromatin immunoprecipitation and tertiary leukemia, as manifested by a significantly shortened followed by next generation sequencing (ChIP-Seq) on sorted latency (Fig. S7A). This shortened latency was not observed in GFP+/YFP+ double positive leukemic cells from primary and the Ezh2-null setting (Fig. S7B). We compared expression pro- secondary AML. Among the polycomb targets, we found genes files of primary and secondary WT and Ezh2-null leukemias. (termed “class 1”) with loss of H3K27me3 upon Ezh2 in- Progression from primary to secondary WT leukemia correlated activation, and polycomb target genes with persistent H3K27me3 with repression of PRC2 module genes (Fig. S7C) and increase despite efficient Ezh2 inactivation (termed “class 2” genes) (Fig. of Myc module genes (Fig. S7E). In contrast, and in keeping with 3A). A class 1 gene set for GSEA was defined as a group of genes the comparable latency of primary and secondary/tertiary AML with significant H3K27me3 (P < 0.01) in wild-type cells and in the Ezh2-null background, there was no significant negative threefold or greater loss of signal at corresponding loci in Ezh2- enrichment of the PRC2 module (Fig. S7D) and no enrichment null leukemic cells. A class 2 gene set was defined as containing of the Myc module (Fig. S7F) observed between Ezh2-null sec- genes with significant H3K27me3 (P < 0.001) in Ezh2-null leu- ondary and Ezh2-null primary leukemia. In summary, these data kemic cells and no decrease in H3K27me3 signal compared with illustrate that Ezh2 is required for leukemia progression to a more the corresponding loci in WT leukemic cells. The gene sets used aggressive disease with shortened latency, an increased fraction of are listed in Dataset S2. actively cycling leukemic cells, and increased organ infiltration. The expression of class 1 genes is enriched in Ezh2-null leu- However, because Ezh2-null cells gave rise to leukemia even in kemic cells compared with WT counterparts as assessed by tertiary recipients, the data also demonstrate that Ezh2 is not GSEA (Fig. 3B). This indicates preferential derepression of strictly required for AML development and self-renewal. polycomb target loci showing loss of H3K27me3 after Ezh2 inactivation. Akin to the situation previously characterized in Genome-Wide H3K27me3 Analysis in Ezh2-Null Leukemic Cells Ezh2-null ES cells (16), we found a group of class 2 genes in Demonstrates Protection from Loss of H3K27me3 on Loci Encoding MLL-AF9 leukemic cells (Dataset S2) with substantial levels of Developmental Regulators. Studies in ES cells have revealed that protected H3K27me3 despite Ezh2 inactivation. These genes are the inactivation of Eed leads to complete loss of H3K27me3, not significantly enriched in Ezh2-null or WT leukemia (Fig. 3B).

Neff et al. PNAS Early Edition | 3of6 Downloaded by guest on September 26, 2021 The class 2 gene set contains many DNA-binding proteins such as Hox-genes, distal-less homeobox genes, paired-box genes, and Gata factors. Using gene-set overlap analysis of this gene set with Gene Ontology gene sets from the Molecular Signatures Data- base (MsigDB) revealed the most significant overlap with anno- tations with a role in development, such as “Anatomical Structure Development,”“Transcription Factor Activity,”“DNA Binding,” and “System Development” (Dataset S2). The preferential activity of the Ezh1-containing PRC2 to de- velopmental regulators is therefore not only observed in ES cells (16) but is also encountered in leukemia. We compared the genes found to be marked by H3K27me3 in Ezh2-null leukemias with data obtained using Ezh2-null murine ES cells. There is highly significant overlap between protected loci in Ezh2-leuke- mic cells and loci both H3K27me3 modified and bound by Ezh1 in Ezh2-null murine ES cells. These data support the concept that Ezh1, the only other H3K27 trimethylase besides Ezh2, partially compensates for the inactivation of Ezh2 in our MLL- AF9 leukemia model (Fig. 3C).

Complete Loss of PRC2 Function Is Incompatible with MLL-AF9 Leukemia. The relatively mild in vivo phenotype of Ezh2 in- activation in MLL-AF9 leukemia was somewhat unexpected. We therefore evaluated the consequences of complete abrogation of PRC2 function. To this end, we used a murine model with floxed sequences of the essential PRC2 component, Eed. We studied the Fig. 4. Eed is strictly required for MLL-AF9 leukemia. (A) Experimental de- Eed sign to test leukemia maintenance: Replated blast colonies derived from effect of inactivation of on the maintenance of established – flox/flox A B Eedflox/flox MLL-AF9 transduced LSK cells from Eed and WT mice were injected leukemia (Fig. 4 and ). Survival of mice in the group into sublethally irradiated primary recipients. The leukemic cells also carried was significantly prolonged compared with recipients of WT cells alleles for MxCre and the YFP-Cre reporter. Leukemic animals were killed (Fig. 4B). Flow cytometric analysis demonstrated consistent out- and secondary sublethally irradiated recipients were injected with sorted growth of GFP single positive cells that had failed to undergo GFP+ cells, (1 × 105 per recipient) harvested from primary recipients. pIpC was deletion of floxed sequences (Fig. 4C). We aimed to further administered to inactivate Eed and to activate the YFP reporter, and the Eed survival of mice was assessed. (B) Significantly prolonged survival in recipi- characterize the effect of inactivation on MLL-AF9 trans- fl fl ents of Eed ox/ ox cells. (C) Flow cytometric analysis demonstrates pre- duced cells. pIpC was administered to primary recipients of fl fl flox/flox dominant outgrowth of unexcised cells. (D) Analysis of Eed ox/ ox cells early Eed and WT MLL-AF9–transduced LSK-derived colonies + (3 d after last dose of pIpC) and late (8 d after last dose of pIpC). Early after and mice were killed 3 or 8 d after the last dose of pIpC. GFP / pIpC, we found efficient deletion in sorted GFP+/YFP+ cells with complete + YFP double positive cells were sorted and analyzed. Deletion of loss of H3K27me3 in three tested leukemias shown here. At the later time Eed sequences was functionally assessed by Western blot for point (8 d after the last dose of pIpC), there was outgrowth of hetero- H3K27me3. Undetectable levels of H3K27me3 were documented zygously inactivated cells with significant levels of H3K27me3 in four of four in three mice (Fig. 4D). Sorted GFP+/YFP+ cells from two of tested leukemias. these mice had minimal residual WT Eed signal by PCR (Fig. S8A) and by qPCR (1.2 and 0.7%). In contrast, all mice assayed these cells were determined to be escapees that had failed to excise 8 d after the last dose of pIpC demonstrated only slightly reduced fl C levels of H3K27me3, consistent with rapid outgrowth of cells that one of the oxed alleles (Fig. S9 ). retain PRC2 function (Fig. 4D). PCR analysis showed bands for Complete Loss of PRC2 Function Accentuates Gene Expression Changes the floxed and the inactivated allele, suggesting outgrowth of cells Compared with Loss of Ezh2. We performed gene expression pro- with heterozygous deletion of floxed Eed sequences (Fig. S8B). filing on Eed-null leukemias with minimal residual WT sequences The phenotype of fully Eed-inactivated leukemias was consistent as assessed by qPCR. Comparing Eed-null to Ezh2-null leukemias with AML (Fig. S8C). To evaluate the capability of Eed-null cells to form leukemia in a noncompetitive setting, we injected sorted (Fig. S10), gene set enrichment analysis demonstrated positive + + enrichment of the PRC2 (Fig. S10A) module and significant GFP /YFP leukemic cells with minimal detectable levels of B H3K27me3 and minimal qPCR signal for WT Eed sequences into negative enrichment of the Myc module (Fig. S10 ). Similar to Ezh2 secondary recipients (Fig. S9A). Recipients of Eed-null cells the situation after inactivation of , decreased expression of Eed showed a significant survival advantage. Four of six recipients of the ES-cell core module was not observed in -null cells (Fig. C Eed-null cells developed leukemia eventually. However, leukemic S10 ). Interestingly, primary MLL-AF9 targets were negatively Eed Ezh2 mice demonstrated outgrowth of incompletely inactivated cells enriched in -null leukemia compared with -null leukemia D Eed (Fig. S9B). We went on to test whether we could demonstrate (Fig. S10 ). Myc protein levels were not decreased in -null successful early engraftment of Eed-null cells in secondary recip- leukemias compared with WT (Fig. S10E). Western blot analysis ients, injecting 1 × 106 sorted live cells per recipient. Whereas WT demonstrated up-regulation of p16ink4a and p19arf in fully Eed- cells showed clear engraftment 18 h, 1 wk, and 1 mo after trans- null primary AML in vivo (Fig. 5B). We then compared de- plantation, very few Eed-null cells engrafted and these cells ex- repression of p16ink4a and p19arf side by side in vitro in WT, panded poorly in vivo (Fig. 5A). Of note, very few Eed-null cells Ezh2-null, and Eed-null cells, 6 d after deletion mediated by were detected in the bone marrow even at 18 h, suggesting not only transduction with a self-excising Cre-encoding vector. Western a proliferative block but also a homing or survival effect of Eed-null blot demonstrates a more pronounced derepression of Cdkn2a leukemic cells in vivo. The maximal engraftment observed in gene products after inactivation of Eed, especially in the case of the Eed-null group was 0.2% in one mouse 1 mo after trans- p19arf (Fig. 5C). Finally, we wanted to analyze the expression of plantation. However, when sorted and subjected to PCR analysis, class 2 genes, which retain significant H3K27me3 in Ezh2-null

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1202258109 Neff et al. Downloaded by guest on September 26, 2021 were able to derive Ezh2-null MLL-AF9 AML up to the third round of serial transplantation. The Ezh2-null leukemias have a lower proportion of actively cycling cells and a lesser degree of organ infiltration. WT leukemia but not Ezh2-null leukemias show significant acceleration upon transplantation from primary into secondary recipients, which is accompanied by transcrip- tional down-regulation of polycomb targets and up-regulation of Myc targets. There is no shortened latency in the secondary and tertiary Ezh2-null MLL-AF9 leukemias, and the enrichment of Myc and PRC2 modules were absent in secondary compared with primary Ezh2-null leukemia. These data show that Ezh2 is required for tumor progression in the MLL-AF9 model. Im- portantly and in contrast to the data presented here, Ezh2 using a conditional knockout model was shown to be essential for lymphomagenesis induced by inactivation of Snf5 (30). This discrepancy demonstrates the importance of context for the role epigenetic modifiers play in cancer. Future studies will correlate the data presented here with data from human leukemia (e.g., at diagnosis and at relapse).

Ezh2-Null Cells Retain Significant H3K27me3. Ezh2 Fig. 5. Detailed analysis of Eed-null MLL-AF9 leukemia. (A) In contrast to Inactivation of WT leukemic cells, Eed-null leukemic cells engraft secondary recipients can be partially compensated in ES cells, where the inactivation poorly. A total of 1 × 106 sorted GFP+/YFP+ cells from primary recipients were of Ezh2 is associated with milder functional defects than the in- transplanted per mouse. n = 3 for each group and time point (except n =2 activation of Eed (16). Recently, it has been suggested that in for WT 1 mo time point). Error bars indicate SEM. (B) Cdkn2a gene products normal hematopoiesis, Ezh1 compensates for inactivation of Ezh2 are up-regulated in Eed-null MLL-AF9 leukemic cells in vivo. (C) Comparison in adult but not fetal cells (31). In murine MLL-AF9 leukemia, of expression of Cdkn2a gene products between sorted preleukemic cells we found a significant number of genes in which H3K27me3 is with WT, Ezh2-null, and Eed-null genotype. LSK-derived MLL-AF9 colonies retained after Ezh2 inactivation. There is large overlap in the set from two separate donors were transduced with self-excising Cre vector and of genes that retain H3K27me3 between ES cells and AML leu- MEDICAL SCIENCES GFP+/YFP+ double positive cells were sorted 6 d after deletion. Note more pronounced derepression of p19arf in Eed-null cells. (D) GSEA analysis of kemic cells, and almost all of these genes are documented Ezh1 class 2 genes, which retain H3K27me3 in Ezh2-null MLL-AF9 AML. This group targets. Ezh1 is the only other known besides of genes is not enriched in Ezh2-null vs. WT leukemia (Fig. 3B) but is dere- Ezh2 to catalyze the trimethylation of H3K27 (16, 32). The ability pressed and significantly enriched in Eed-null vs. Ezh2-null leukemic cells. of Ezh1 to compensate for Ezh2 inactivation in cancer appears to be context dependent (30). Little is currently known about the composition and recruitment of PRC2-Ezh1 in cancer. Our data leukemia and are not enriched in Ezh2-null compared with WT suggest a role for Ezh1 in MLL-AF9 leukemia that will need to be leukemia. This gene set is highly significantly enriched in Eed-null addressed in more detail in future studies. AML compared with Ezh2-null leukemia (Fig. 5D). In summary, inactivation of Eed leads to complete loss of PRC2 function and Eed Is Absolutely Required for MLL-AF9 Leukemia. We evaluated the complete loss of H3K27me3. This loss is accompanied by de- response of MLL-AF9 leukemia to complete loss of PRC2 repression of polycomb targets silenced by persistent H3K27me3 function by inactivation of the essential component, Eed. This is in Ezh2-null leukemic cells and by attenuated expression of a Myc accompanied by more profound gene-expression changes than module with documented importance in cancer. These gene ex- after inactivation of Ezh2. Specifically, a subset of genes retain- pression changes are accompanied by failure of Eed-null leukemic ing highly significant H3K27me3 after inactivation of Ezh2 is cells to proliferate in vivo, demonstrating that in contrast to murine derepressed after inactivation of Eed, but not after inactivation ES cells (16, 26) and murine epidermis (27), MLL-AF9 leukemia of Ezh2. Complete loss of PRC2 function after inactivation of strictly depends on PRC2 function. Eed is incompatible with leukemia growth. This is in contrast to ES cells and the early embryo, where growth after complete loss Discussion of PRC2 function is compromised, but not abolished (16, 26, 33). Conflicting data exist regarding the role of PRC2 in cancer. El- PRC2 function has also been shown to be dispensable for pro- evated levels of PRC2 components have been shown to be as- liferation of Ezh1/Ezh2 double-knockout cells in the epidermis sociated with aggressiveness, invasiveness, and survival in solid proper but required in the hair follicle (27). In summary, avail- tumors (5). A gain-of-function mutation for EZH2 has been able data illustrate that the consequences of partial and com- described in large cell lymphoma (28). On the other hand, het- plete loss of PRC2 function are context dependent. erozygous and homozygous mutation and/or deletion of EZH2 It is noteworthy that two transcriptional modules derived from has been described in some hematologic malignancies (8–12, 29), the study of ES cells (20) are found in a leukemia model, and but is not typically seen in AML. Understanding of PRC2 biology that they appear functionally linked. Our data suggest that in cancer is limited by the heterogeneity of clinical samples and modulation of polycomb function may be a strategy to thera- the pleiotropy of genetic programs controlled by PRC2. We here peutically decrease expression of Myc transcriptional targets, at present data using a genetically defined murine model to study least in MLL-AF9 leukemia. In this respect it is reassuring that the effects of homozygous inactivation of genes encoding the Ezh2 does not appear required for adult normal hematopoietic PRC2 components Ezh2 and Eed on MLL-AF9 leukemia. stem cell function (31). There is little doubt that a cure for MLL- AF9 AML will require combination therapies. The model de- Ezh2 Is Not Strictly Required for MLL-AF9 AML. A requirement for scribed here will allow us to further study pathways that collab- Ezh2 in malignant self-renewal in murine cancer models has orate with PRC2 in the maintenance of MLL-rearranged been suggested, either using shRNA approaches (14) or geneti- leukemia, with the goal of defining combinations of critical drug cally engineered ablation (30). In contrast, we provide evidence targets and of eventually developing rational drug combinations that Ezh2 is not strictly required for MLL-AF9 leukemia, as we to better treat MLL-AF9 AML.

Neff et al. PNAS Early Edition | 5of6 Downloaded by guest on September 26, 2021 Materials and Methods modification. Briefly, cross-linking was performed with 1% formalin, and the For antibodies and detailed experimental procedures, please refer to SI cells were lysed in SDS buffer. DNA was fragmented by sonication and his- Materials and Methods. tones and bound DNA were precipitated overnight. Eluted DNA fragments were analyzed by qPCR or subjected to sequencing using next-generation Solexa sequencing. Mice and Breeding. Animals were maintained at the animal research facility at Children’s Hospital Boston. Animal experiments were approved by the fi × 5– institutional animal care and use committee. Conditional mice were main- RNA Ampli cation and Gene Expression Array. RNA was isolated from 1 10 × 5 + + fi tained on a C57B/6 background crossed to MxCre and ROSA26-flox-STOP- 5 10 sorted GFP /YFP cells using TRIzol (Invitrogen), ampli ed (Ovation flox-EYFP mice. Pico WTA; NuGEN), labeled (NuGEN; Encore Biotin Module), and hybridized to Affymetrix 430_2 murine microarrays. Generation of Transformed Murine Cells and Leukemia. Ecotropic retroviral vectors were generated by cotransfection of 293T cells with packaging Data Analysis and Statistical Methods. Precipitated DNA from ChIP experi- constructs. LSK cells were transduced with MLL-AF9-GFP and maintained in ments was sequenced on the Illumina HiSeq 2000 platform. Reads that methylcellulose with supplemental cytokines. In some experiments, cells were aligned to multiple loci in the genome were discarded. The ChIP-Seq signal fi then transduced with hit-and-run Cre and 5–6 d later, GFP+/YFP+ cells were was quanti ed as total number of reads per million in the region 1 kb up- sorted and plated. Replated colonies were transplanted into C57B/6 synge- stream of the transcription start site (TSS) to 1 kb downstream of TSS for neic sublethally irradiated (600 rad) recipients at 1–2 × 105 cells per mouse. H3K27me3. An empirical background distribution model of reads was con- For secondary transplants, GFP+/YFP+ sorted cells from leukemic mice were structed to find the significance level of signal at a gene. Expression array data transplanted. were analyzed with GenePattern release 3.2.3–3.3.3 (http://www.broadinstitute. org/cancer/software/genepattern/), and the probe sets were mapped to genes fi Biochemical Assays: Cell Growth, Apoptosis, Cell Cycle Analysis, Western using library les (version 30) downloaded from the Affymetrix Website. Blotting, and qPCR. Cell cycle analysis was performed using the BrdU-APC/ GSEA was performed using www.broadinstitute.org/gsea. The integrative fi 7AAD kit from BD PharMingen. BrdU was administered i.p. to leukemic mice analysis of histone modi cation levels and gene expression was performed and mice were killed and analyzed 1 h later. Antibodies used are detailed in SI using iCanPlot. Statistical tests were performed using GraphPad Prism5; for Materials and Methods. For colony assays, sorted transduced leukemia cells details see SI Materials and Methods. were plated in methylcellulose M3234 (Stem Cell Technologies) containing IL3, IL6, and murine stem cell factor (mSCF) at 1,000–5,000 cells per plate, and ACKNOWLEDGMENTS. We thank Kathrin Bernt, Adrian Bracken, Demetrios replated every 5–7d.Insomeexperiments,GFP+/YFP+ were sorted before the Kalaitzidis, and Andre Krivtsov for helpful discussions and Ron Mathieu fl first and second platings. for expert advice regarding ow cytometry and FACS. This project was supported by National Cancer Institute (NCI) Grant 1K08CA154777-01 and

+ + the Amgen Hematology and Oncology Fellowship Stipend (to T.N.); NCI ChIP-qPCR and ChIP-Seq. GFP /YFP leukemic cells were isolated by FACS Grants P01CA66996, U01CA105423, and P30DK049216; funding from the sorting. For ChIP of H3K27me3, ChIP was performed as previously described American Cancer Society and the Leukemia and Lymphoma Society (to S.A.A.); (34), using a polyclonal rabbit antibody (Millipore; 07449) specifictothe and funding from the Howard Hughes Medical Institute (to S.H.O.).

1. Boyer LA, et al. (2006) Polycomb complexes repress developmental regulators in 19. Somervaille TC, et al. (2009) Hierarchical maintenance of MLL myeloid leukemia stem murine embryonic stem cells. Nature 441:349–353. cells employs a transcriptional program shared with embryonic rather than adult stem 2. Lee TI, et al. (2006) Control of developmental regulators by Polycomb in human cells. Cell Stem Cell 4:129–140. embryonic stem cells. Cell 125:301–313. 20. Kim J, et al. (2010) A Myc network accounts for similarities between embryonic stem 3. Morey L, Helin K (2010) Polycomb group protein-mediated repression of transcrip- and cancer cell transcription programs. Cell 143:313–324. tion. Trends Biochem Sci 35:323–332. 21. Bernt KM, et al. (2011) MLL-rearranged leukemia is dependent on aberrant H3K79 4. Cao R, et al. (2002) Role of lysine 27 methylation in Polycomb-group si- methylation by DOT1L. Cancer Cell 20:66–78. lencing. Science 298:1039–1043. 22. Delmore JE, et al. (2011) BET bromodomain inhibition as a therapeutic strategy to 5. Varambally S, et al. (2002) The polycomb group protein EZH2 is involved in pro- target c-Myc. Cell 146:904–917. – gression of prostate cancer. Nature 419:624 629. 23. Zuber J, et al. (2011) RNAi screen identifies Brd4 as a therapeutic target in acute 6. Kamminga LM, et al. (2006) The Polycomb group gene Ezh2 prevents hematopoietic myeloid leukaemia. Nature 478:524–528. – stem cell exhaustion. Blood 107:2170 2179. 24. Dawson MA, et al. (2011) Inhibition of BET recruitment to chromatin as an effective 7. Croonquist PA, Van Ness B (2005) The polycomb group protein enhancer of zeste treatment for MLL-fusion leukaemia. Nature 478:529–533. fl homolog 2 (EZH 2) is an oncogene that in uences myeloma cell growth and the 25. Neri F, et al. (2012) Myc regulates the transcription of PRC2 to control the expression mutant ras phenotype. Oncogene 24:6269–6280. of developmental genes in embryonic stem cells. Mol Cell Biol 32:840–851. 8. Guglielmelli P, et al. (2011) EZH2 mutational status predicts poor survival in myelo- 26. Chamberlain SJ, Yee D, Magnuson T (2008) Polycomb repressive complex 2 is dis- fibrosis. Blood 118:5227–5234. pensable for maintenance of embryonic stem cell pluripotency. Stem Cells 26:1496– 9. Ernst T, et al. (2010) Inactivating mutations of the histone methyltransferase gene 1505. EZH2 in myeloid disorders. Nat Genet 42:722–726. 27. Ezhkova E, et al. (2011) EZH1 and EZH2 cogovern histone H3K27 trimethylation and 10. Nikoloski G, et al. (2010) Somatic mutations of the histone methyltransferase gene are essential for hair follicle homeostasis and wound repair. Genes Dev 25:485–498. EZH2 in myelodysplastic syndromes. Nat Genet 42:665–667. 28. Sneeringer CJ, et al. (2010) Coordinated activities of wild-type plus mutant EZH2 drive 11. Ntziachristos P, et al. (2012) Genetic inactivation of the polycomb repressive complex tumor-associated hypertrimethylation of lysine 27 on histone H3 (H3K27) in human B- 2 in T cell acute lymphoblastic leukemia. Nat Med 18:298–303. cell lymphomas. Proc Natl Acad Sci USA 107:20980–20985. 12. Zhang J, et al. (2012) The genetic basis of early T-cell precursor acute lymphoblastic 29. Jankowska AM, et al. (2011) Mutational spectrum analysis of chronic myelomonocytic leukaemia. Nature 481:157–163. 13. Martinez-Garcia E, Licht JD (2010) Deregulation of H3K27 methylation in cancer. Nat leukemia includes genes associated with epigenetic regulation: UTX, EZH2, and – Genet 42:100–101. DNMT3A. Blood 118:3932 3941. 14. Suvà ML, et al. (2009) EZH2 is essential for glioblastoma cancer stem cell maintenance. 30. Wilson BG, et al. (2010) Epigenetic antagonism between polycomb and SWI/SNF – Cancer Res 69:9211–9218. complexes during oncogenic transformation. Cancer Cell 18:316 328. 15. Cheng AS, et al. (2011) EZH2-mediated concordant repression of Wnt antagonists 31. Mochizuki-Kashio M, et al. (2011) Dependency on the polycomb gene Ezh2 dis- promotes β-catenin-dependent hepatocarcinogenesis. Cancer Res 71:4028–4039. tinguishes fetal from adult hematopoietic stem cells. Blood 118:6553–6561. 16. Shen X, et al. (2008) EZH1 mediates methylation on histone H3 lysine 27 and com- 32. Margueron R, et al. (2008) Ezh1 and Ezh2 maintain repressive chromatin through plements EZH2 in maintaining stem cell identity and executing pluripotency. Mol Cell different mechanisms. Mol Cell 32:503–518. 32:491–502. 33. Pasini D, Bracken AP, Hansen JB, Capillo M, Helin K (2007) The polycomb group 17. Srinivas S, et al. (2001) Cre reporter strains produced by targeted insertion of EYFP protein Suz12 is required for embryonic stem cell differentiation. Mol Cell Biol 27: and ECFP into the ROSA26 locus. BMC Dev Biol 1:4. 3769–3779. 18. Silver DP, Livingston DM (2001) Self-excising retroviral vectors encoding the Cre re- 34. Krivtsov AV, et al. (2008) H3K79 methylation profiles define murine and human combinase overcome Cre-mediated cellular toxicity. Mol Cell 8:233–243. MLL-AF4 leukemias. Cancer Cell 14:355–368.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1202258109 Neff et al. Downloaded by guest on September 26, 2021