Modeling and analysis of acute leukemia using hematopoietic stem and progenitor cells A dissertation submitted to the Graduate School of the University of Cincinnati in partial fulfillment of the requirement for the degree of Doctor of Philosophy in the Molecular &Developmental Biology Graduate Program of the College of Medicine by Shan Lin

BS, Tsinghua University, 2009

Dissertation Committee: James C. Mulloy, PhD (Chair) Geraldine Guasch, PhD Ashish R. Kumar, MD, PhD Daniel T. Starczynowski, PhD Yi Zheng, PhD Abstract

For decades, elegant models have yielded important insights into the complex biology of acute leukemia development. However, species differences between human and mouse could have significant influences on biological and translational applications. Therefore, human primary hematopoietic cells and xenograft mouse models have become important research tools in the field. In this report, we briefly review the methodologies that use human primary hematopoietic cells to model acute leukemia and examine the effects of leukemic oncogenes. The advantages and limitations of the human model system compared to syngeneic mouse models are discussed. The species-related complexity in human disease modeling is highlighted in the study establishing a faithful model of proB-ALL caused by MLL-AF4, the fusion product of the t(4;11). MLL-AF4 proB-ALL has poor prognosis, the lack of an accurate model hampers the study of disease pathobiology and therapeutic testing. We find human AF4 cDNA inhibits retroviral production and efficient transduction, this limitation can be overcome by fusing MLL with murine Af4, highly conserved with human AF4. Whereas MLL-Af4-transduced murine cells induce only

AML, transduced human CD34+ cells produce proB-ALL faithful to t(4;11) disease, fully recapitulating the immunophenotypic and molecular aspects of the disease. We find that MLL-fusion leukemia shows genetic heterogeneity driven by differential DNA binding of the fusion . We report lineage plasticity as a new mechanism of resistance to CD19-directed therapy in t(4;11) patients that exemplifies the clinic relevance of our model. This model can provide unique insight for targeting t(4;11) ALL. In addition, by using a pre-leukemia model of AML1-ETO, the fusion generated by t(8;21) and associated with AML, we identify FOXO1 as an essential self-renewal factor in this disease. FOXO1 is consistently upregulated in t(8;21) AML and functions as a critical oncogenic mediator rather than a tumor suppressor. Expression of FOXO1 in human CD34+ cells promotes a pre-leukemic state, partially phenocopying the cellular and transcriptional effects of AML1-ETO expression. These functions are specific to FOXO1 and are not elicited by FOXO3, and the DNA binding ability of FOXO1 is required.

AML1-ETO and FOXO1 co-occupy the majority of their binding sites, whereby FOXO1 binds to

i multiple crucial self-renewal and is necessary for their activation. In agreement with this observation, loss of FOXO1 inhibits the long-term proliferation and clonogenicity of AML1-ETO cells.

Thus, increased FOXO1 represents a new mechanism for acquiring aberrant self-renewal by pre-leukemia stem cells and provides a novel target for therapeutic intervention. This study also serves as an example to demonstrate the usefulness and capacity of the human model system to dissect critical targets and pathways for acute leukemia development.

ii

iii

Acknowledgements

I would like to thank my advisor Dr. James Mulloy. It is really fortunate for a graduate student to have a great mentor as you. Thank you for providing me with constant support not only for research work but also helping solve many non-scientific issues, so that I can focus on science, which is the more interesting part of the Ph.D life. You are always optimistic and encouraging, leading me to see the promising things during the struggling, and to overcome all the difficulties. Thank you for providing me with an open environment to develop scientifically and independently.

I also want to thank other committee members Drs. Yi Zheng, Ashish Kumar, Daniel Starczynowski and

Guasch Geraldine for their invaluable advice and patience through my doctoral training. They are always ready for help with broad knowledge and critical thinking, which benefit me a lot in the pursuit of my graduate degree and future studies. Thanks the Division of Experimental Hematology and Biology for having such a nice collaborating atmosphere, everybody is willing to help. I appreciate Drs. Lee

Grime, Gang Huang, Mohammad Azam and other faculty members, whose brilliant ideas and suggestions have broadened my view of my own projects as well as leukemia research. Huge Thanks to our wonderful collaborators, Dr. Michael Thirman and his colleagues in University of Chicago, and Dr. Constanze

Bonifer and her group from University of Birmingham.

Thanks all current and previous members in Mulloy lab, I would not be able to finish my study without their selfless support and hard work. Kevin gave me a great orientation when I joined the lab, preparing me with all necessary techniques for leukemia research. He, Susumu and Fu-sheng spent a lot of their private time to revise my proposal for qualifier examination, that is why I can survive. Mark, Christina and Mahesh have taken care most of my mice experiments. Andy and Janet make really delicious dessert.

Ben and Eric, like other people, never say no when I need a help. Also to all visiting scholars and students and supporting staff that I have worked with, Shuhong, Vase, Dino, Alba, Sam, Rushi, Bargav, Ann,

Courtney, Mahima, Navin and Ryan, thank you for making lab fun and I appreciate all I learn from you.

Special thanks to Alba for hard work to gather data for me even when she was in a stressful period. And iv thanks rotating students Jordan and David for giving me an opportunity to be a mentor, and your contributions really help my project, I hope you will make a great success for your Ph.D study. I would also thank colleagues from other labs, with you I can quickly learn new skills and get my experiments cooking.

Last but not least are my dear family, without whom I would never become who I am now. Many thanks to my dad and mom, Xiaohui Lin and Manli Peng, who are my greatest supporters. Even though you do not know too much about the research, your unwavering moral support and understanding carry me through all difficulties and challenges of the doctoral training, and your life experiences will also be invaluable reference for my future life. Thanks to my lovely girlfriend, Liu Yang, you are the sunshine, wiping out all my stress and depression. Thanks all friends, Bo Liu, Xuan Zhou, Sha Wang, Qin Wang,

Rui Wang, MukYan Wong, Kuan-shen Wong, Jiang Jiang Yang, Zhen Zhang, Jieqing Fan, Jia You,

Cindy Whitaker, Shaun Rudolph and Tamala Whitaker, who help me to settle down in Cincinnati, from renting an apartment to fixing a car, you are always there when I am in trouble.

Thanks for all the people whose names are not mentioned due to limited page space but whose help will never be forgot.

v

Contents

List of figures……………………………………………………………………………………...1

Chapter 1. General introduction…………………………………………………………………...2

1.1 Acute leukemia…………………………………………………………………………2

1.2 Mouse model for leukemia study………………………………………………………….2

1.3 Human cell model system for acute leukemia study……………………….……………....7

1.4 Limitation of the human cell model system…………………………………………...11

1.5 References.…………………………………………………………………………..14

Chapter 2. Establishment of a faithful MLL-AF4 proB-ALL model using human cells………...19

2.1 Introduction………………………………………………………………………………19

2.2 Results………...….………………………………………………………………………29

2.3 Discussion………....………...…………………………………………………………..44

2.4 Experimental procedures…….…………………………………………………………...48

2.5 References………………………………………………………………………………..55

2.6 Appendix Tables….………………………………………………………………………61

Chapter 3. FOXO1-induced self-renewal defines the AML1-ETO pre-leukemic program………68

3.1 Introduction………………………………………………………………………………68

3.2 Results….…………………………………………………...……………………………76

3.3 Discussion………………………………………………………………………………..87

3.4 Experimental procedures…….…………………………………………………………...90

3.5 Reference………………………………………………………………………………....97

Chapter 4. Future implications………………………………………………………………….103

vi

4.1 Future directions………………………………………………………………………...103

4.2 References………………………………………………………………………………106

vii

List of figures

Figure 1-1 Experimental models for leukemia study………………………………..………………………3

Figure 2-1 Schematic representation of the MLL protein and MLL fusions……………………………….20

Figure 2-2 Stably expressed MLL-Af4 in mouse HSPCs induces AML…………………………………...31

Figure 2-3 Human CD34+ cells expressing MLL-Af4 initiate proB ALL in NSG mice…………………...33

Figure 2-4 MLL-Af4 proB-ALL is transplantable…………………………………………………………34

Figure 2-5 signatures derived from t(4;11) patients are enriched in MLL-Af4 proB-ALL…………..35

Figure 2-6 MLL-Af4 and MLL-AF9 lead to block at distinct B cell development stage…………………...37

Figure 2-7 MLL-Af4 profile recapitulates t(4;11) specific molecular signature………….39

Figure 2-8 MLL-fusions promote distinct gene expression profiles via differential DNA binding………...40

Figure 2-9. MLL-Af4 maintains lymphoid potential after myeloid priming……………………………….41

Figure 2-10 MLL-Af4 phenotypic myeloid cells keep an active lymphoid program………………………42

Figure 2-11 Phenotypic flexibility of t(4;11) ALL contributes to resistance to CD19-taregeted therapy…..45

Figure 3-1 t(8;21) AML is associated with increased FOXO1……………………………………………..76

Figure 3-2 FOXO1 is upregulated in AE pre-leukemia cells……………………………………………….77

Figure 3-3 FOXO1 inhibition impairs growth of AE cells…………………………………………………79

Figure 3-4 Increased FOXO1 promotes a pre-leukemic state in human CD34+ HSPCs…………………...81

Figure 3-5 FOXO3 does not display the oncogenic function………………………………………………82

Figure 3-6 ROS reduction cannot recapitulate the effects of increased FOXO1…………………………...83

Figure 3-7 Transcriptome of FOXO1 recapitulates gene signature of AE………………………………….85

Figure 3-8 FOXO1 regulates AE key target genes…………………………………………………………86

1

Chapter 1 General introduction

1.1 Acute leukemia

Acute leukemia is a clonal hematopoietic malignancy caused by genetic mutations that disrupt normal hematopoiesis, leading to the accumulation of abnormal hematopoietic cells with excessive proliferation and impaired differentiation. Acute leukemia is a biologically and genetically heterogeneous disease.

Based on the immunophenotype and hematopoietic lineage markers manifested, the disease can be generally characterized as (AML) or acute lymphocytic leukemia (ALL). The

American Cancer Society estimated that 13,800 cases of AML and 6000 cases of ALL were diagnosed in the United States in 2012, and only 50-70% of patients survived[1]. Recent research has shown leukemogenesis is a stepwise progression, with initiating genetic mutations establishing pre-leukemia stem cells that evolve to leukemia stem cells (LSC) after receiving additional driver mutations. The LSC can self-renew and generate all other bulk leukemia cells and is the cell population that sustains the disease [2]. Numerous genetic mutations associating with acute leukemia have been identified, including chromosomal translocations that account for approximately half of the cases. translocation usually results in generating fusion proteins involving regulators critical for normal hematopoiesis, thus the fusion proteins can interfere with the normal function and alter the transcription network of the native regulators, promoting leukemogenesis. In the following sections, we will focus on two types of chromosome translocations in particular: the t(4;11)(q21;q23) which fuses MLL(KMT2A) to

AF4 (AFF1) and accounts for 10% of ALL[3]; and the t(8;21)(q22;q22) that is present in 10% of AML and brings AML1 (RUNX1) and ETO (RUNX1T1) together[4].

1.2 Mouse model for leukemia study

For decades, the syngeneic mouse has been the most popular genetic tool for modeling leukemia and investigating the mechanistic basis of leukemogenesis. Both viral transduction -

2 transplantation system and genetically engineered mice have been used by many investigators (Figures 1-

1 A and B).

Figure 1-1. Experimental models for leukemia study. (A) Mouse bone marrow viral transduction - transplantation system. Mouse HSPC are isolated from donor BM and transduced with retro- or lenti- which overexpress a gene or a shRNA targeting the endogenous genes. The transduced cells are transplanted into recipient mice and their leukemogenesis capacity is analyzed. (B) Genetically engineered mouse model. Leukemia-related genes are manipulated genetically in murine fertilized ovum or embryonic stem cells, which are then used to generate the transgenic mouse line. (C) Human cell model system. Similar to (A), human HSPC or leukemia cells are isolated from healthy donor or patients, transduced by , and transplanted into immunodeficient mice to evaluate the normal or malignant hematopoiesis.

3

For instance, to understand the function of AML1-ETO (AE), mouse hematopoietic stem and progenitor cells (HSPC) have been transduced with AE. Overexpression of AE expanded the stem cell compartment and modestly impaired myeloid differentiation, resulting in the accumulation of myeloid progenitors within the mouse bone marrow (BM), but was insufficient to induce leukemia[5]. The similar phenotype was observed in an inducible AE transgenic mouse model where AE expression was regulated by a tet-off system[6]. A conditional AE knock-in mouse model using the inducible Mx1-Cre has also been generated. Interestingly, though enhanced self-renewal of myeloid progenitors was observed, AE did not block differentiation in this knock-in model[7]. This result suggests that, when not driven by the endogenous , AE could be expressed at a supraphysiological level in retroviral transduction and transgenic mouse models, which causes differentiation block. Yet it could also indicate that achieving a high expression of oncogene is a critical step in AE leukemogenesis, as it was reported that AE abundance is much lower in remission than at the time of diagnosis in patients[8]. Nevertheless, all these

AE mouse models reveal that AE by itself can only promote the HSPC towards a pre-leukemia state, but is not capable to induce the overt disease. This is in accordance with the observation that the HSPC bearing t(8;21) translocation can be detected years before disease onset, suggesting a pre-leukemic stage is present in patients[9]. Recent genome mutation profiling has shown that most t(8;21) patients have at least one additional genetic abnormality, which potentially contributes to the disease development[10,

11]. Indeed, in mouse models, it has been demonstrated that introducing additional genetic hits along with

AE can successfully induce AML[12, 13]. In addition, a recent study using mouse models to screen for cooperating events for AE AML was able to identify functional mutations that also occur in t(8;21) patients[14]. These examples demonstrate that mouse models can generally mimic the human disease phenotype and therefore provide useful insights for understanding the crucial pathways in leukemogenesis and developing therapy.

However, application of knowledge gained from the mouse model is not without caveats. First of all, while a relative homogeneous genetic background can improve the consistency of the experiment results,

4 the inbred nature of the syngeneic mouse raises concern that the phenotype could be mouse strain specific. It has been reported that the role of the cell cycle regulator p21(Cdkn1a) in mouse hematopoietic stem cells (HSC) depends on genetic background. The lack of p21 impaired the self-renewal of HSC in mice on the C57BL/6 and 129 mixed background, in contrast, it seems that p21 is not essential for HSC of mice with pure C57BL/6 background[15, 16]. Thus it could be more complicated to extrapolate results from mice to with more distal and diverse genetic backgrounds.

Secondly, despite that the general roadmap and basic biology of hematopoiesis are largely conserved between mouse and human, significant species specificity exists and cannot be ignored. Accurate immunophenotypic characterization of stem cells and progenitors in normal and malignant hematopoiesis is essential for identifying the cell-of-origin of leukemia, the dysregulated molecular pathways involved in self-renewal and differentiation and the therapeutic targets for eradicating LSC. Current evidence shows that surface marker expression on human HSC have little correspondence with murine HSC.

Murine HSC are enriched in a Lineage-Sca-1+c-Kit+CD150+CD48- population of BM[17]. In contrast,

Sca-1, CD150 and CD48 are not useful markers for identification of human HSC, since there is no human homolog of Sca-1, and CD150 and CD48 are not differentially expressed between human HSC and progenitors[18]. On the other hand, it has been shown that human HSC reside in the CD34+CD38-

THY1+CD45RA-CD49f+ compartment[19], while murine HSC do not express CD34 and do express

CD38[20]. Accordingly, although the immunophenotype of human LSC has flexibility and variety, CD34 and CD38 are useful markers for isolating LSC in patients[21]. Instead, c-Kit expression labels the LSC in mouse leukemia models[22, 23]. Therefore, it could be difficult to choose the proper cell population in human corresponding to those studied in mouse leukemia models, impeding the of the findings. Difference of intrinsic biological properties beyond cell surface marker also exist between human and murine HSC. Flt3 ligand is a potent factor maintaining the survival of human HSC in vitro[24], in contrast, Flt-3 signaling is important for lymphopoiesis but not for HSC survival in mice as murine HSC do not express Flt-3[25, 26]. As another example, HoxB4 is a

5 critical self-renewal regulator of murine HSC, the overexpression of which induced a 1000 fold expansion of murine HSC[27]. However, only around 2 fold expansion can be achieved when HoxB4 was expressed in human HSC[28, 29]. These different results demonstrate that the discoveries from mouse studies could either miss the critical targets or lose the fidelity in human context.

Additionally, protein–protein interactions and related signal transduction pathways that exist in humans may be absent or different in mice, whereby expression of human leukemia associated oncogenes in murine cells may not properly reproduce the human disease. Of note, although researchers have tried for many years, the mouse model faithfully recapitulating t(4;11) MLL-AF4 ALL has not been available.

Clinically, MLL-AF4 is almost exclusively associated with ALL with B cell lineage (B-ALL), accounting for 50-70% of infant ALL (< 1 year) and around 10% of B-ALL of all ages. MLL-AF4 ALL has a special pro-B cell immunophenotype with CD19+CD34+ but lacking CD10 expression. While overall B-ALL patients have a >70% survival rate, MLL-AF4 B-ALL has an extremely poor prognosis, with a 5-year survival rate of only 30%[3, 30]. Due to its special clinical features and aggressiveness, MLL-AF4 ALL has drawn lots of attention in laboratories. Mouse models using various approaches to express the fusion protein have been established. In a transgenic mouse model, where the human derived MLL-AF4 gene was driven by a MSCV promoter, B cell lymphoma was found after 14 months of latency[31]. The first knock-in mice was built by homologous recombination, where human AF4 was knocked into the mouse

Mll . But these mice developed lymphoid and myeloid hyperplasia and B-cell lymphomas at 22 months of age[32]. The second knock-in mouse was created by using the invertor conditional technology, where a floxed AF4 cDNA was knocked into Mll opposite to the transcription direction, thus Mll-AF4 could only be expressed in the presence of Cre. Similarly, these mice developed mature B-cell neoplasia as oppose to pro-B ALL[33]. A later knock-in model used Mx1-Cre and a Lox-Stop-Lox cassette to express Mll-AF4 conditionally. Although half of the mice developed AML, ALL were successfully induced in the other half of the mice. However, most ALL cases were characterized as pre- or common B-

ALL, in which the leukemic cells were blocked at a later development stage than pro-B. In contrast, pro-B

6

ALL only consisted of 5% of all leukemia cases[34]. Therefore, there are limitations in modeling acute leukemia using mouse cells. In order to overcome these difficulties and with the concern that species and cell specificity may hamper the clinical translation of the research, more and more people have begun to use human primary HSPC to model and study acute leukemia.

1.3 Human cell model system for acute leukemia study

The experimental strategy manipulating human cells is similar to the retroviral transduction and transplantation approach (Figure 1-1 C). The CD34+ or lineage negative HSPC are isolated from human cord blood cells or mobilized BM cells. It is now recognized that cells of different etiology differ in their properties, including self-renewal capacity and lineage preference[35], thus the source of the HSPC could affect the propensity to be transformed and alter the disease phenotype[36]. The widely used way to introduce oncogenes into human HSPC is via retroviral or lentiviral delivery systems. Recently, people have reported the utilization of site-specific nucleases TALEN-mediated genome editing to create chromosome translocations involving the MLL gene in CD34+ cells[37]. We expect the genome editing by new generation technology CRISPR/Cas9 will be applied to human cell leukemia modeling soon[38].

The liquid culture and methylcellulose culture systems have been developed that allow expansion of normal or manipulated human HSPC and investigation of their progenitor function and differentiation capacity in vitro[39], while it is owing to the continual improvement of the immunodeficient mouse based xenograft system that we can interrogate how an oncogene may contribute to leukemogenesis in vivo.

Through years of development in the community, the most widely used immunodeficient mouse strain so far is based on the NOD-scid-IL2Rgnull mouse. Depending on whether it is a truncation or a deletion in the

IL-2R , the strains are named as NOG and NSG, respectively[40, 41]. The mice with spontaneous severe combined immunodeficiency mutation (scid) harbor disruption in Prkdc gene, which plays an essential role in repair of DNA double strand break [42]. Prkdc is required for B and T cells development which goes through DNA recombination processes[43]. Thus the scid mice lack the adaptive arm of the . Nonobese diabetic (NOD) mice have dampened innate immunity

7 against human cells due to harboring a polymorphism in Sirpa gene. SIRPa is a highly polymorphic transmembrane protein expressed on myeloid cells[44]. CD47, the ligand of SIRPa, functions as a “don't eat me” signal, and its interaction with SIRPa results in inhibition of phagocytosis by [45].

Although CD47 is ubiquitously expressed on human hematopoietic cells, it generally does not interact with murine SIRPa, thus cannot protect the human cells from mouse macrophages. However, the Sirpa polymorphism of NOD mice allows binding of human CD47 with high affinity and thereby induces host tolerance to human engrafted cells[46]. Enlighted by this, mice engineered to express human

SIRPa molecule has also been used to avoid phagocytosis[47]. IL-2R common gamma chain is a critical component for IL-2, IL-4, IL-7, IL-9, IL-15, and IL-21 signaling, and thus for lymphocyte development.

Loss of this gene in mice results in further loss of B and T cells, and ablation of NK cells[40, 41].

Thereby, the NOG and NSG mice have severe defects in both adaptive and innate immunity, supporting high level engraftment of human cells. Since the Prkdc gene is expressed broadly in many tissues, scid mice are more sensitive and vulnerable to treatment inducing DNA damage, such as irradiation and chemotherapy reagents, which is unfavorable for using these mice in pre-clinical studies. To improve this aspect, people have developed another way to diminish B and T cells by deleting recombination activating genes 1 or 2 (Rag1 and Rag2), which are involved in DNA recombination of T and B cell genes, and expressed exclusively in hematopoietic cells. As the rest of the body harbors an intact

DNA repair system, Ragnull mice show a much better tolerance to DNA damaging agents than scid mice[48]. It has been shown that NRG mice, in which the Rag1 deletion is introduced to the NOD-IL2r-/- background, can sustain an equivalent level of human engraftment as NSG mice[49]. Therefore, NRG mice can be a useful strain for leukemia studies related to therapy development.

Ablation of mature mouse cell populations is usually required to make niche space for human engrafted cells. Thus, preconditioning of the recipient mice is commonly performed before transplantation. One prevailing preconditioning approach is by sublethal γ-irradiation. Treating the mice with chemotherapeutic agents, such as busulfan, is another widely used method[50, 51]. Additionally, it has

8 been reported that the selective specifically recognizing and clearing mouse cells but not human cells can be used for preconditioning[52]. The engraftment efficiency of human malignant cells correlates to the aggressiveness of the disease[53], hence preconditioning sometimes is not required for aggressive leukemia samples. This is due to the acquired advantage of proliferation and self-renewal of these leukemia cells which enhance their competence over mouse cells. Since the preconditioning may cause the disease unrelated death of the mice at a small percentage, especially in NSG mice as discussed above, whether to precondition needs to be determined according to experimental design. There are several ways available to transplant human cells into recipient mice. The intravenous and intra-bone injections are common approaches when using adult recipients. The BM niche provides critical factors to support the survival and proliferation of HSPC. Due to the fact that most of the human and mouse BM homing and retention signals are cross reactive[54], the human HSPC can home to mouse BM through intravenous injection, which is a relatively easier procedure than other transplantation approaches. Although whether the BM is absolutely required for LSC is still uncertain, it has been shown that human LSC can reside within the BM endosteal region in NSG mice[55]. Thus in general, intravenous injection is also sufficient for efficient BM engraftment of human leukemia cells. However, for the leukemia samples of limited cell number or with homing difficulties, direct injection into the bone is preferred. Injection into the liver of newborn mice is an alternative strategy for reliable engraftment [56]. In addition, enhancing the homing signaling of the transplanted cells can also improve BM engraftment[57].

More and more efforts have been put into building leukemia models using primary human HSPC, and the effects of many individual leukemia-associated oncogenes on human HSPC have been examined.

Nevertheless, most oncogenes tested elicit only partial phenotypes without overt disease. This is consistent with the concept that multiple genetic insults are required for stepwise progression to transformation in leukemia development. In contrast, the strong transformation capacity has been observed using MLL-rearrangement fusion proteins. Our group and others have shown that both ALL and

AML can be induced in immunodeficient mice transplanted with human CD34+HSPC expressing MLL-

9

AF9 or MLL-ENL, the fusion products of t(9;11)(p22;q23) and t(11;19)(q23;p13), respectively[58, 59], suggesting that MLL-fusion proteins are sufficient to transform human HSPC by themselves. This is in agreement with the leukemia genome sequencing analysis that revealed that MLL- rearrangement leukemia patients, especially of young age, have a silent genome without many additional mutations[60,

61]. As a proof of the stepwise transformation model, people have shown introducing additional genetic hits can generate a full leukemia. For example, while human cells expressing t(9;22)(q34;q11) fusion product BCL-ABL alone cannot initiate leukemia, B-ALL is observed when co-expressing BMI1 with

BCL-ABL[62]. And although TLS-ERG derived from the t(16;21) only induced a pre-leukemia phenotype in human cells by itself, TLS-ERG pre-leukemic clones were able to evolve into full leukemia eventually after accumulating other mutations[63]. In spite of these successes, human cells overall are refractory to transformation, in stark contrast to mouse cells. While murine leukemia was successfully developed upon overexpression of oncogenes such as HOXA10, mutated RAS and NUP98-HOXA9[64-

66], these oncoproteins can only enhance the self-renewal with a myeloid proliferative phenotype in human cells[67-69]. Especially, an AE leukemia model derived from human HSPC is not yet available even when a cooperating mutation including KIT and RAS mutant was introduced[70, 71], although both have been demonstrated to work successfully with AE in mouse system[13, 72]. This distinct resistance to transformation between human and murine cells has also been reported in other non-hematopoietic tissues[73], which could reflect that human cells intrinsically have evolved to impose more safety mechanisms for tumor suppression, likely due to humans face increased risks of mutation accumulation and transformation compared with mice because of longer lifespan. Interestingly, it seems that human and murine HSC manifest different cellular behavior in response to DNA damage. Irradiation-induced DNA damage in murine HSC activates a dependent non-homologous end joining (NHEJ) DNA repair response, which is in favor of maintaining cell survival but may introduce mutations due to imperfect repair[74]. In contrast, irradiation activates a p53 dependent apoptotic response in human HSC[75], suggesting that human HSC with DNA damage are sacrificed in exchange for genome stability. This and other protective mechanisms may also be activated upon expression of oncogenes, and need to be 10 overcome during the progression to transformation. For instance, in human AE pre-leukemia cells, it was reported that the KIT mutation can cooperate to enhance cell proliferation by improving DNA repair and reducing DNA damage and induced by AE[71]. In addition to the intrinsic cellular difference,

BM microenvironment also plays a critical role in supporting leukemia initiation and progression, and thus the leukemic readout of human cells is dependent on the microenvironment provided by immunodeficient mice. However the murine BM niche is not optimal for all types of human leukemia cells (discussed later), therefore the lack of transformation by oncogenes could be a false-negative due to lack of a supportive microenvironment. Nonetheless, differential resistance to transformation between

HSPC from the two species have important implications for the application of results of mouse studies in understanding molecular pathway accounting for human leukemogenesis.

1.4 Limitation of the human cell model system

No experimental model system is perfect, and some limitations of the human cell model need to be kept in mind during experimental design. As caveats of using virus to deliver genetic alterations in other studies, retroviral integration into the genome may cause nonspecific gene disruption or activation. In addition, gene dosage has been shown as a crucial factor determining the phenotype, multiple copies of integrated virus can result in the expression of transduced gene at a supraphysiological level, leading to artificial phenotypes[76]. Therefore, the virus should always be titered. Many virus vectors are designed to co-express a selected fluorescent or cell surface marker, the expression of which correlates with the dosage of the transduced genes, thus allowing us to purifying cells expressing the target gene at a proper level according to the intensity of markers. In addition, genome editing tools now enable the use of an endogenous promoter to drive an oncogene, avoiding the gene dosage issue. As mentioned in the previous section, multiple cooperating genetic hits are required for leukemia development. Although the genome editing technology introduces the flexibility to manipulate the genome in human cells, generating compound mutations is still much more difficult than in the mouse system, in which it can be easily

11 achieved by cross breeding of different transgenic mouse lines. Further advances in genome editing will provide solutions to facilitate introducing compound mutations in human HSPC.

One additional caveat to xenograft leukemia modeling is the lack of a functional immune system in recipient mice. Until now, the full spectrum of human hematopoiesis is unable to be fully recapitulated in immunodeficient mice. In humans, myeloid cells comprise more than half of total white blood cells, however this portion is reduced to 10-20% in human engraftment of NSG mice[77]. Proper human erythropoiesis and thrombopoiesis are not supported in immunodeficient mice, the mature human erythrocytes and are almost completely absent in xenografted mice[49]. Not only the full lineage potential of human HSPC is restricted, function defects have also been reported in B, T and myeloid lineage developed in mice to various extent[77, 78]. Thus, the evaluation of lineage potential of human

HSPC and differentiation block caused by oncogene relies more on in vitro assays, which also have caveats in that these do not represent the actual physiological conditions. Recently, studies have shown that infection and inflammation play important roles in initiation and progression of leukemia[79, 80].

However, since the xenograft immune system is defective natively and cannot be completely reconstituted by human cells, studying the relationship between leukemogenesis and infection, inflammation and other immune system responses in immunodeficient mice will be difficult. The same reason also hampers usage of the xenograft system for preclinical study of immunotherapy that releases the check point of immune system, in which the endogenous immune cells are boosted to attack tumor cells[81].

Although the capacity of immunodeficient mice to host leukemia samples has been greatly improved, a substantial number of samples, especially of less aggressive disease, still fail to engraft. For example, robust xenograftment of t(8;21) patient samples has not been feasible, placing a block on translating findings from the model system to preclinical testing on patient samples. The inability to sustain complete human hematopoiesis or engraftment of leukemia samples indicates that the mouse BM niche lacks some crucial growth factors, potentially from the low level of conservation and thus poor cross-reactivity of some cytokines between human and mouse. For instance, human and mouse IL-3 proteins, a critical

12 cytokine for survival and proliferation of stem cells and myeloid progenitors, only show conservation in

30% of the sequence and do not have inter-species reactivity. Several approaches have been utilized to construct a more humanized and functional niche in mice. Human BM derived stroma cells and/or human cytokines can be co-transplanted with human hematopoietic cells, which has shown efficacy to improve the efficiency and function of engraftment[82]. Since the stroma cells and cytokine proteins cannot last long in vivo, the effects are usually temporary, or multiple rounds of transplantation are required to provide long-term benefits. Alternatively, people have engineered the mouse genome to express human cytokines necessary for human hematopoiesis. NSG mice transgenic for cytokines that support human myelopoiesis (human SCF, GM-CSF and IL3, NSGS) allow increased myeloid output of human HSPC and better engraftment of AML samples compared to control NSG mice[83]. The transgenic approach raises the concern that the cytokines are expressed at a non-physiological level and may lead to unexpected consequence on engrafted cells[84]. Thus, human cytokine genes knocked into the corresponding mouse locus is considered as a better approach. The new generation of immune deficient strain MISTRG mice have been developed, which express human M-CSF, IL-3, GM-CSF,

Thrombopoietin (THPO) as knock-ins and human SIRPa as transgene. Although more studies are needed, the current data suggest MISTRG mice show improved host ability for human HSPC and malignant cells, enhanced myelopoiesis, and development of functional NK cells. However, the human erythropoiesis and thrombopoiesis defects are still not corrected in these mice[85].

Though with several limitations, the human cell model system provides a platform of leukemia study that is relevant to clinic studies, allowing us to compare and evaluate the data generated from the model with those from the patients directly. It is a useful tool that is complementary, sometimes even superior, to mouse genetic systems in leukemia research. In the following sections, two projects will be presented to demonstrate the power and capacity of the human cell model system facilitating our understanding of acute leukemia. In Chapter 2, we report the establishment of a faithful MLL-AF4 pro-B ALL leukemia model using human HSPC and xenograft that has not been feasible in the mouse system. In Chapter 3, we

13 identified FOXO1 as the oncogenic mediator critical for aberrant self-renewal of the AE pre-leukemia cells, using a human AE pre-leukemia model. Future implications are discussed in Chapter 4.

1.5 References

1. Rose-Inman, H. and D. Kuehl, Acute leukemia. Emerg Med Clin North Am, 2014. 32(3): p. 579-96. 2. Jan, M. and R. Majeti, Clonal evolution of acute leukemia genomes. Oncogene, 2013. 32(2): p. 135-40. 3. Marchesi, F., K. Girardi, and G. Avvisati, Pathogenetic, Clinical, and Prognostic Features of Adult t(4;11)(q21;q23)/MLL-AF4 Positive B-Cell Acute Lymphoblastic Leukemia. Adv Hematol, 2011. 2011: p. 621627. 4. Muller, A.M., et al., Complementing mutations in leukemias: from mouse models to clinical applications. Oncogene, 2008. 27(44): p. 5759-73. 5. de Guzman, C.G., et al., expansion and distinct myeloid developmental abnormalities in a murine model of the AML1-ETO translocation. Mol Cell Biol, 2002. 22(15): p. 5506-17. 6. Rhoades, K.L., et al., Analysis of the role of AML1-ETO in leukemogenesis, using an inducible transgenic mouse model. Blood, 2000. 96(6): p. 2108-15. 7. Higuchi, M., et al., Expression of a conditional AML1-ETO oncogene bypasses embryonic lethality and establishes a murine model of human t(8;21) acute myeloid leukemia. Cancer Cell, 2002. 1(1): p. 63-74. 8. Shima, T., et al., The ordered acquisition of Class II and Class I mutations directs formation of human t(8;21) acute myelogenous leukemia stem cell. Exp Hematol, 2014. 42(11): p. 955-65 e1- 5. 9. Wiemels, J.L., et al., In utero origin of t(8;21) AML1-ETO translocations in childhood acute myeloid leukemia. Blood, 2002. 99(10): p. 3801-5. 10. Krauth, M.T., et al., High number of additional genetic lesions in acute myeloid leukemia with t(8;21)/RUNX1-RUNX1T1: frequency and impact on clinical outcome. Leukemia, 2014. 28(7): p. 1449-58. 11. Micol, J.B., et al., Frequent ASXL2 mutations in acute myeloid leukemia patients with t(8;21)/RUNX1-RUNX1T1 chromosomal translocations. Blood, 2014. 124(9): p. 1445-9. 12. Schessl, C., et al., The AML1-ETO fusion gene and the FLT3 length mutation collaborate in inducing acute leukemia in mice. J Clin Invest, 2005. 115(8): p. 2159-68. 13. Wang, Y.Y., et al., C-KIT mutation cooperates with full-length AML1-ETO to induce acute myeloid leukemia in mice. Proc Natl Acad Sci U S A, 2011. 108(6): p. 2450-5. 14. Hatlen, M.A., et al., Integrative genetic analysis of mouse and human AML identifies cooperating disease alleles. J Exp Med, 2016. 213(1): p. 25-34. 15. Cheng, T., et al., Hematopoietic stem cell quiescence maintained by p21cip1/waf1. Science, 2000. 287(5459): p. 1804-8. 16. van Os, R., et al., A Limited role for p21Cip1/Waf1 in maintaining normal hematopoietic stem cell functioning. Stem Cells, 2007. 25(4): p. 836-43. 17. Kiel, M.J., et al., SLAM family receptors distinguish hematopoietic stem and progenitor cells and reveal endothelial niches for stem cells. Cell, 2005. 121(7): p. 1109-21. 18. Sintes, J., et al., Differential expression of CD150 (SLAM) family receptors by human hematopoietic stem and progenitor cells. Exp Hematol, 2008. 36(9): p. 1199-204. 19. Notta, F., et al., Isolation of single human hematopoietic stem cells capable of long-term multilineage engraftment. Science, 2011. 333(6039): p. 218-21.

14

20. Randall, T.D., et al., Expression of murine CD38 defines a population of long-term reconstituting hematopoietic stem cells. Blood, 1996. 87(10): p. 4057-67. 21. Goardon, N., et al., Coexistence of LMPP-like and GMP-like leukemia stem cells in acute myeloid leukemia. Cancer Cell, 2011. 19(1): p. 138-52. 22. Somervaille, T.C. and M.L. Cleary, Identification and characterization of leukemia stem cells in murine MLL-AF9 acute myeloid leukemia. Cancer Cell, 2006. 10(4): p. 257-68. 23. Krivtsov, A.V., et al., Transformation from committed progenitor to leukaemia stem cell initiated by MLL-AF9. Nature, 2006. 442(7104): p. 818-22. 24. Sitnicka, E., et al., Human CD34+ hematopoietic stem cells capable of multilineage engrafting NOD/SCID mice express flt3: distinct flt3 and c- expression and response patterns on mouse and candidate human hematopoietic stem cells. Blood, 2003. 102(3): p. 881-6. 25. Adolfsson, J., et al., Upregulation of Flt3 expression within the bone marrow Lin(-)Sca1(+)c-kit(+) stem cell compartment is accompanied by loss of self-renewal capacity. Immunity, 2001. 15(4): p. 659-69. 26. Sitnicka, E., et al., Key role of flt3 ligand in regulation of the common lymphoid progenitor but not in maintenance of the hematopoietic stem cell pool. Immunity, 2002. 17(4): p. 463-72. 27. Antonchuk, J., G. Sauvageau, and R.K. Humphries, HOXB4-induced expansion of adult hematopoietic stem cells ex vivo. Cell, 2002. 109(1): p. 39-45. 28. Buske, C., et al., Deregulated expression of HOXB4 enhances the primitive growth activity of human hematopoietic cells. Blood, 2002. 100(3): p. 862-8. 29. Amsellem, S., et al., Ex vivo expansion of human hematopoietic stem cells by direct delivery of the HOXB4 homeoprotein. Nat Med, 2003. 9(11): p. 1423-7. 30. Pui, C.H., M.V. Relling, and J.R. Downing, Acute lymphoblastic leukemia. N Engl J Med, 2004. 350(15): p. 1535-48. 31. Tamai, H., et al., Activated K-Ras protein accelerates human MLL/AF4-induced leukemo- lymphomogenicity in a transgenic mouse model. Leukemia, 2011. 25(5): p. 888-91. 32. Chen, W., et al., A murine Mll-AF4 knock-in model results in lymphoid and myeloid deregulation and hematologic malignancy. Blood, 2006. 108(2): p. 669-77. 33. Metzler, M., et al., A conditional model of MLL-AF4 B-cell tumourigenesis using invertor technology. Oncogene, 2006. 25(22): p. 3093-103. 34. Krivtsov, A.V., et al., H3K79 Methylation Profiles Define Murine and Human MLL-AF4 Leukemias. Cancer Cell, 2008. 14: p. 355-368. 35. Babovic, S. and C.J. Eaves, Hierarchical organization of fetal and adult hematopoietic stem cells. Exp Cell Res, 2014. 329(2): p. 185-91. 36. Horton, S.J., et al., MLL-AF9-mediated immortalization of human hematopoietic cells along different lineages changes during ontogeny. Leukemia, 2013. 27: p. 1116-26. 37. Buechele, C., et al., MLL leukemia induction by genome editing of human CD34+ hematopoietic cells. Blood, 2015. 126(14): p. 1683-94. 38. Hendel, A., et al., Chemically modified guide enhance CRISPR-Cas genome editing in human primary cells. Nat Biotechnol, 2015. 33(9): p. 985-9. 39. Notta, F., et al., Distinct routes of lineage development reshape the human blood hierarchy across ontogeny. Science, 2016. 351(6269): p. aab2116. 40. Ito, M., et al., NOD/SCID/gamma(c)(null) mouse: an excellent recipient mouse model for engraftment of human cells. Blood, 2002. 100(9): p. 3175-82. 41. Shultz, L.D., et al., Human lymphoid and myeloid cell development in NOD/LtSz-scid IL2R gamma null mice engrafted with mobilized human hemopoietic stem cells. J Immunol, 2005. 174(10): p. 6477-89.

15

42. Blunt, T., et al., Identification of a nonsense mutation in the carboxyl-terminal region of DNA- dependent protein catalytic subunit in the scid mouse. Proc Natl Acad Sci U S A, 1996. 93(19): p. 10285-90. 43. van der Burg, M., J.J. van Dongen, and D.C. van Gent, DNA-PKcs deficiency in human: long predicted, finally found. Curr Opin Allergy Clin Immunol, 2009. 9(6): p. 503-9. 44. Adams, S., et al., Signal-regulatory protein is selectively expressed by myeloid and neuronal cells. J Immunol, 1998. 161(4): p. 1853-9. 45. Tsai, R.K. and D.E. Discher, Inhibition of "self" engulfment through deactivation of myosin-II at the phagocytic synapse between human cells. J Cell Biol, 2008. 180(5): p. 989-1003. 46. Takenaka, K., et al., Polymorphism in Sirpa modulates engraftment of human hematopoietic stem cells. Nat Immunol, 2007. 8(12): p. 1313-23. 47. Strowig, T., et al., Transgenic expression of human signal regulatory protein alpha in Rag2-/- gamma(c)-/- mice improves engraftment of human hematopoietic cells in humanized mice. Proc Natl Acad Sci U S A, 2011. 108(32): p. 13218-23. 48. Shultz, L.D., F. Ishikawa, and D.L. Greiner, Humanized mice in translational biomedical research. Nat Rev Immunol, 2007. 7(2): p. 118-30. 49. Pearson, T., et al., Non-obese diabetic-recombination activating gene-1 (NOD-Rag1 null) interleukin (IL)-2 receptor common gamma chain (IL2r gamma null) null mice: a radioresistant model for human lymphohaematopoietic engraftment. Clin Exp Immunol, 2008. 154(2): p. 270- 84. 50. Wunderlich, M., et al., OKT3 prevents xenogeneic GVHD and allows reliable xenograft initiation from unfractionated human hematopoietic tissues. Blood, 2014. 123(24): p. e134-44. 51. Hayakawa, J., et al., Busulfan produces efficient human cell engraftment in NOD/LtSz-Scid IL2Rgamma(null) mice. Stem Cells, 2009. 27(1): p. 175-82. 52. Czechowicz, A., et al., Efficient transplantation via antibody-based clearance of hematopoietic stem cell niches. Science, 2007. 318(5854): p. 1296-9. 53. Pearce, D.J., et al., AML engraftment in the NOD/SCID assay reflects the outcome of AML: implications for our understanding of the heterogeneity of AML. Blood, 2006. 107(3): p. 1166-73. 54. Peled, A., et al., Dependence of human stem cell engraftment and repopulation of NOD/SCID mice on CXCR4. Science, 1999. 283(5403): p. 845-8. 55. Saito, Y., et al., Identification of therapeutic targets for quiescent, chemotherapy-resistant human leukemia stem cells. Sci Transl Med, 2010. 2(17): p. 17ra9. 56. Gimeno, R., et al., Monitoring the effect of gene silencing by RNA interference in human CD34+ cells injected into newborn RAG2-/- gammac-/- mice: functional inactivation of p53 in developing T cells. Blood, 2004. 104(13): p. 3886-93. 57. Brault, L., et al., CXCR4-SERINE339 regulates cellular adhesion, retention and mobilization, and is a marker for poor prognosis in acute myeloid leukemia. Leukemia, 2014. 28(3): p. 566-76. 58. Barabe, F., et al., Modeling the initiation and progression of human acute leukemia in mice. Science, 2007. 316(5824): p. 600-4. 59. Wei, J., et al., Microenvironment Determines Lineage Fate in a Human Model of MLL-AF9 Leukemia. Cancer Cell, 2008. 13: p. 483-495. 60. Dobbins, S.E., et al., The silent mutational landscape of infant MLL-AF4 pro-B acute lymphoblastic leukemia. Genes Cancer, 2013. 52(10): p. 954-60. 61. Andersson, A.K., et al., The landscape of somatic mutations in infant MLL-rearranged acute lymphoblastic leukemias. Nat Genet, 2015. 47: p. 330-337. 62. Rizo, A., et al., BMI1 collaborates with BCR-ABL in leukemic transformation of human CD34+ cells. Blood, 2010. 116(22): p. 4621-30.

16

63. Warner, J.K., et al., Direct evidence for cooperating genetic events in the leukemic transformation of normal human hematopoietic cells. Leukemia, 2005. 19(10): p. 1794-805. 64. Thorsteinsdottir, U., et al., Overexpression of HOXA10 in murine hematopoietic cells perturbs both myeloid and lymphoid differentiation and leads to acute myeloid leukemia. Mol Cell Biol, 1997. 17(1): p. 495-505. 65. Parikh, C., R. Subrahmanyam, and R. Ren, Oncogenic NRAS, KRAS, and HRAS exhibit different leukemogenic potentials in mice. Cancer Res, 2007. 67(15): p. 7139-46. 66. Calvo, K.R., et al., Nup98-HoxA9 immortalizes myeloid progenitors, enforces expression of Hoxa9, Hoxa7 and Meis1, and alters cytokine-specific responses in a manner similar to that induced by retroviral co-expression of Hoxa9 and Meis1. Oncogene, 2002. 21(27): p. 4247-56. 67. Buske, C., et al., Overexpression of HOXA10 perturbs human lymphomyelopoiesis in vitro and in vivo. Blood, 2001. 97(8): p. 2286-92. 68. Shen, S.W., et al., Mutant N-ras preferentially drives human CD34+ hematopoietic progenitor cells into myeloid differentiation and proliferation both in vitro and in the NOD/SCID mouse. Exp Hematol, 2004. 32(9): p. 852-60. 69. Chung, K.Y., et al., Enforced expression of NUP98-HOXA9 in human CD34(+) cells enhances stem cell proliferation. Cancer Res, 2006. 66(24): p. 11781-91. 70. Chou, F.S., et al., N-Ras(G12D) induces features of stepwise transformation in preleukemic human umbilical cord blood cultures expressing the AML1-ETO fusion gene. Blood, 2011. 117(7): p. 2237-40. 71. Wichmann, C., et al., Activating c-KIT mutations confer oncogenic cooperativity and rescue RUNX1/ETO-induced DNA damage and apoptosis in human primary CD34+ hematopoietic progenitors. Leukemia, 2015. 29(2): p. 279-89. 72. Zhao, S., et al., KRAS (G12D) cooperates with AML1/ETO to initiate a mouse model mimicking human acute myeloid leukemia. Cell Physiol Biochem, 2014. 33(1): p. 78-87. 73. Rangarajan, A., et al., Species- and cell type-specific requirements for cellular transformation. Cancer Cell, 2004. 6(2): p. 171-83. 74. Mohrin, M., et al., Hematopoietic stem cell quiescence promotes error-prone DNA repair and mutagenesis. Cell Stem Cell, 2010. 7(2): p. 174-85. 75. Milyavsky, M., et al., A distinctive DNA damage response in human hematopoietic stem cells reveals an apoptosis-independent role for p53 in self-renewal. Cell Stem Cell, 2010. 7(2): p. 186- 97. 76. Baum, C., et al., Side effects of retroviral gene transfer into hematopoietic stem cells. Blood, 2003. 101(6): p. 2099-114. 77. Ishikawa, F., et al., Development of functional human blood and immune systems in NOD/SCID/IL2 receptor {gamma} chain(null) mice. Blood, 2005. 106(5): p. 1565-73. 78. Gille, C., et al., Monocytes derived from humanized neonatal NOD/SCID/IL2Rgamma(null) mice are phenotypically immature and exhibit functional impairments. Hum Immunol, 2012. 73(4): p. 346-54. 79. Greaves, M., Infection, immune responses and the aetiology of childhood leukaemia. Nat Rev Cancer, 2006. 6(3): p. 193-203. 80. Martin-Lorenzo, A., et al., Infection Exposure is a Causal Factor in B-cell Precursor Acute Lymphoblastic Leukemia as a Result of Pax5-Inherited Susceptibility. Cancer Discov, 2015. 5(12): p. 1328-43. 81. Hasegawa, K., et al., An Immunocompetent Mouse Model for MLL/AF9 Leukemia Reveals the Potential of Spontaneous Cytotoxic T-Cell Response to an Antigen Expressed in Leukemia Cells. PLoS One, 2015. 10(12): p. e0144594.

17

82. Sontakke, P., et al., Modeling BCR-ABL and MLL-AF9 leukemia in a human bone marrow-like scaffold-based xenograft model. Leukemia, 2016. 83. Wunderlich, M., et al., AML xenograft efficiency is significantly improved in NOD/SCID-IL2RG mice constitutively expressing human SCF, GM-CSF and IL-3. Leukemia, 2010. 24(10): p. 1785-8. 84. Nicolini, F.E., et al., NOD/SCID mice engineered to express human IL-3, GM-CSF and Steel factor constitutively mobilize engrafted human progenitors and compromise human stem cell regeneration. Leukemia, 2004. 18(2): p. 341-7. 85. Rongvaux, A., et al., Development and function of human innate immune cells in a humanized mouse model. Nat Biotechnol, 2014. 32(4): p. 364-72.

18

Chapter 2 Establishment of a faithful MLL-AF4 proB-ALL model using human cells

2.1 Introduction

MLL and its role in normal hematopoiesis

The mixed‐lineage leukemia gene (MLL, also called KMT2A), the mammalian homologue of Drosophila trithorax gene, encodes a large multi-domain protein involving in gene regulation (Figure 2-1). The maturation of the MLL protein undergoes a site‐specific cleavage by ‐aspartase Taspase‐1, which generates an amino terminal MLL (MLL-N), and a carboxyl terminal MLL (MLL-C) fragments

[1]. The MLL-N contains several domains mediating protein-protein and protein-chromatin interactions, and a conserved SET domain with histone H3 4 (H3K4) methyltransferase activity is identified within the MLL-C[2, 3]. The two fragments of MLL interact via the FYRN domain on MLL-N and the

FYRC and SET domains on MLL-C[4]. It was reported that the intra-molecular interaction helped the

MLL protein to be stabilized and fulfill full function, whereas the proteolytic cleavage is dispensable[5].

Trithorax mutants show homeotic transformations in drosophila, the phenotype of which is similar to mutations in homeobox (Hox) genes. Indeed, trithorax has been identified as a crucial activator for

Hox[6]. Several groups have established Mll knockout mice, confirming MLL is also an important Hox regulator in . Deletion of both Mll alleles is embryonic lethal, and mice died between E12.5 to

E16.5 in different reports[7, 8], while Mll+/- mouse embryos can survive but manifest homeotic transformations of the axial skeleton[9]. Reduced expression of Hox genes has been confirmed in cells from Mll-deficient mice[9, 10]. Analysis of fetal liver from Mll-deficient embryos showed defects in the

HSC, as Mll-deficient fetal liver cells have reduced colony forming units (CFU) and fail to repopulate recipient mice when transplanted. Inability of Mll-/- HSC to maintain quiescence was thought as one factor contributing to the phenotype[7, 8]. Conditional knockout mice have been made to study Mll function in postnatal hematopoiesis. The pan-haematopoietic Vav-Cre was used to induce gene excision around E13.5, avoiding embryonic lethality. In one report, Vav-Cre mediated Mll deletion did not result in deficiency in adult homeostatic hematopoiesis, however the HSC from these mice lost the ability in 19 competitive reconstitution assays[8]. However, another group reported that not only was there a reconstitution defect, the conditional deletion of Mll also caused a dramatic reduction in HSPC numbers at steady state, followed by multilineage defects such as anemia and thrombocytopenia. Only half of those mice can survive past 3 weeks of age[11]. The discrepancy of two studies is likely due to the unique strategies used to generated the Mll loss-of-function allele by excising different , thus leading to distinct in-frame deletion within MLL-N protein. These results suggest that each domain of MLL protein could have independent functions contributing to normal hematopoiesis. Given that Hox genes have profound effects on proliferation and differentiation of HSPC, deregulation of Hox gene expression is also thought to account for hematopoietic abnormalities caused by Mll deficiency. It was shown that reintroduction of Hox genes into Mll-/- embryoid bodies can rescue their impaired hematopoietic development potential[12]. In addition to Hox genes, other Mll targets critical for maintaining mouse HSC molecular programs were identified by microarray profiling, including Prdm16 and Evi1[13].

20

Figure 2-1. Schematic representation of the MLL protein and MLL fusions. The mixed lineage leukaemia (MLL) gene encodes a nuclear protein with complex domain structure (featured domains are highlighted). The mature MLL protein consists of two non-covalently associated subunits MLL-N and MLL-C, which is produced by Taspase1 mediated cleavage ( two cleavage sites indicated by single arrow) and associate via FYRN and FYRC domains. Important protein interactions are indicated by double arrows. The N terminus of MLL interacts with MENIN and LEDGF, which is crucial for chromatin target recognition. The CXXC motif within the RD1 domain mediates DNA binding through associating with non-methylated CpGs. In addition, interacting with polymerase associated factor complex (PAFc) also plays a role in recruitment of MLL to its target genes. A typical MLL fusion protein contains the MLL-N to the PHD domain and the C terminus is replaced by one of over 70 fusion partner genes. Through various mechanisms, the fusion partner proteins can recruit DOT1L complex or super elongation complex (SEC), leading to dysregulated transcription elongation and aberrant gene activation, and thus acute leukemia. RD, repression domain. PHD, homeodomain. BD, bromodomain.

In flies, the SET domain and the histone methyltransferase activity of trithorax is critical for maintaining activation of Hox genes[6]. This requirement seems conserved in mouse, as mice with SET domain deletion of MLL show reduced H3K4 monomethylation as well as Hox gene expression in whole embryo and embryonic fibroblasts. However, unlike other germline Mll deficient mice, these mice can survive into adulthood, although they have skeletal abnormalities as expected[14]. Interestingly, there are no hematopoietic defects observed in homozygous SET deletion mice, and accordingly, SET deletion has no impact on Hox gene expression or H3K4 methylation in hematopoietic cells. This result suggests that

MLL-associated MOF histone acetyltransferase, possessing H4K16 acetylation activity, plays the dominant role in maintaining MLL targets in mouse HSPC[15]. Therefore, the requirement for methyltransferase activity by MLL on gene regulation is likely cell context-dependent.

The functions of MLL-N domains in gene regulation have also been studied, revealing their critical roles in recruiting MLL protein to its target gene loci (Figure 2-1). The N terminus of MLL-N interacts with

MENIN, which in turn tethers to LEDGF[16, 17]. LEDGF contains a PWWP domain, which binds to nucleosome and recognizes histone H3 di/tri-methyl-lysine 36 (H3K36me2/3)[18, 19]. H3K36me3 is associated with actively transcribed gene bodies, while H3K36me2 is slightly more enriched in the promoter proximal region[20]. In accordance, MLL was reported to associate with promoter regions

21 preferentially, but the MLL binding sites also exist in gene bodies[21]. However, by using Menin knockout mice, researchers showed that expression of some MLL target genes were not altered by loss of

Menin, suggesting MLL also regulates MENIN-independent pathways[22]. Adjacent to the N terminus are three AT‐hooks and two speckled nuclear localization (SNL) signals. It has been suggested that the

AT‐hooks and the SNL signals confer the sub‐nuclear localization of MLL[23], deletion of which prevents MLL from nuclear entry[7]. Two repression domains (RDs), which can recruit corepressors such as HDAC1 and BMI1, have been identified by luciferase reporter assays[24]. The CXXC motif residing in RD1 binds to non-methylated CpGs[25]. An enrichment of non-methylated CpGs in the promoters of

MLL target genes has been revealed accordingly[18]. In addition, polymerase associated factor complex

(PAFc) interacts with a region spanning CXXC and RD2[26]. The PAFc is an RNA polymerase II (Pol

II)-associated complex with a role in H3K4me3 maintenance and transcription activation[27]. It was shown that binding to PAFc is necessary for recruitment of MLL to HoxA9 locus in MEF cells[28].

Therefore, it is likely that MLL tends to associate with chromatin having features of active promoters, and thus has a function in maintaining the active transcriptional state of target genes. However, since MLL does not have sequence specific DNA-binding activity, how the specificity of its targets is achieved is not well understood.

MLL-rearrangement (MLL-r) acute leukemia

As its name indicates, human MLL was initially identified as the gene disrupted in both acute myeloid and lymphoid leukaemias associating with the chromosomal translocations at 11q23. MLL-r accounts for

~10-15% of total acute leukemia cases and particularly ~70-90% of infant ALL[29]. MLL-r is considered as a poor prognosis factor in patients with ALL. MLL-r ALL patients have a high risk of relapse and an estimated 5-year event-free survival (EFS) rates of 30%, which is much lower than the >80% EFS of

ALL without MLL-r. MLL-r is regarded as an intermediate risk factor in AML, with EFS of ~30-

50%[30].

22

The breakpoint cluster region (BCR) spans exons 8-13 of MLL gene, thus a typical MLL fusion protein contains the N terminus of MLL proximal to the PHD domains, and the C terminus of fusion partners[29].

Therefore, those domains mediating wildtype MLL-chromatin interactions are retained in MLL-fusions and are necessary for their gene regulation (Figure 2-1). Loss of Menin or Ledgf results in a reduction of colony formation in murine cells transformed by MLL-ENL[17]. In addition, overexpressing an interfering peptide derived from the MLL N terminus or LEDGF interacting region delays MLL-AF9 induced AML[31, 32]. The CXXC motif and its DNA binding capacity are required for MLL-ENL transformation[33]. Association with PAFc is also necessary for tethering MLL-fusion proteins to the

HoxA9 locus[26]. Furthermore, knockdown of PAFc components or disruption the PAFc interaction with an interfering peptide impairs colony forming ability of MLL-AF9 cells[34]. Strikingly, it was demonstrated that a chimeric protein fusing the PWWP domain of LEDGF with the minimum CXXC motif and ENL is sufficient to induce AML using mouse bone marrow cells[18], emphasizing the functional significance of these complexes.

Since the MLL-C portion is not retained in MLL-fusion proteins, it appears that the transcription regulation of the target genes relies on the fusion partners. So far more than 70 MLL translocation partners have been identified, yet ~80% of MLL-r involve AF4 (AFF1), ENL (MLLT1), AF9 (MLLT3),

AF6 (MLLT4), AF10 (MLLT10) and ELL[35]. Although the functions of many other rare fusion partners contributing to leukemogenesis are less clear, several mechanisms by which the frequent fusion partners dysregulate target genes are revealed through years of studies. AF4, ENL, AF9 and ELL have all been identified to reside in a purified complex named super elongation complex (SEC), which also contains

AF5q31(AFF4), ELL-associated factor (EAF) and complex of positive transcription elongation factor-b

(PTEF-b)[36]. Of note, AF5q31 is also a rare MLL partner[37]. ELL was demonstrated to stimulate the transcription elongation rate of Pol II [38]. P-TEFb is a Pol II CTD kinase comprised of the kinase CDK9 and its regulatory subunit, Cyclin T1 or Cyclin T2. It has an essential role in the release of poised Pol II for elongation by active of the RNA Pol II CTD as well as inhibitory phosphorylation of

23

NELF and DSIF, the negative elongation factors[39]. Therefore, one mechanism whereby MLL-fusion proteins cause aberrant gene activation is through recruitment of SEC and stimulating elongation in an unregulated manner. Knockdown of the SEC components such as AF5q31 and ENL have been shown to reduce the transformation capacity of MLL-fusions[36, 40]. In addition to SEC, AF9 and ENL are involved in another complex with the H3K79 methyltransferase DOT1L, which also includes AF10 and another rare MLL fusion protein AF17 (MLLT6)[41]. The DOT1L catalyzes H3K79 methylation.

Although H3K79me2 correlates with transcriptionally active genes in general[42], data suggests that

DOT1L recruited by MLL-fusions results in ectopic H3K79me2 of target genes which is critical for the gene activation[43, 44]. While the mechanism through which increased H3K79me2 leads to gene activation requires further study, it has been shown that genetic ablation of DOT1L or AF10 resulted in the loss of MLL-AF9 transforming ability[45, 46], and small molecules that inhibit DOT1L enzymatic activity have efficacy in inhibition of MLLr leukemia cell growth[47]. Interestingly, several studies have shown that DOT1L is not associating with SEC components besides AF9 and ENL, and that AF9 and

ENL interact with DOT1L complex in a manner that is mutually exclusive with SEC[36, 48, 49]. More complexity comes from the study showing that AF9 and ENL exist in separate SEC complexes displaying similar but non-identical functions[48], which is also the case for AF4 and AF5q31[50]. Since most of these studies were performed in non-hematopoietic cells such as 293 and Hela cells, and involved the overexpression of wildtype fusion partners, whether these interactions exist with MLL-fusion proteins within leukemia cells requires validation. Nevertheless, these data suggest that one MLL-fusion could recruit multiple complexes for gene regulation, and different types of MLL-fusions may associate with distinct complexes and thus regulate targets in different ways.

Dimerization by the fusion partner portion serves as another mechanism of MLL-fusion mediated gene dysregulation. MLL-fusions including MLL-AF6 and MLL-GAS7 are thought to use this strategy, as the dimerization domains in the fusion partners have been shown to be crucial for their transforming abilities[51, 52]. Surprisingly, MLL artificially fused to a dimerization domain was able to transform

24 mouse HSPC in a dimerization dependent manner[51]. The mechanism by which the dimerization causes gene dysregulation is not understood. Intriguingly, although AF6 does not interact with either the SEC or

DOT1L complex directly, MLL-AF6 leukemia is still sensitive to the disruption of these complexes[40,

53], suggesting that the dimerization might generate a platform for SEC and DOT1L recruitment.

As a summary of previous findings, a general model of MLL-r mediated leukemogenesis has been proposed. Retaining chromatin binding domains of wildtype MLL, MLL-fusion proteins are tethered to

MLL binding sites, which then recruit SEC and DOT1L complexes leading to the uncontrolled activation of MLL target genes, among which the HOXA genes and their co-factor MEIS1 are the best-known.

Indeed, the hyper-activation of HOXA and MEIS1 genes are observed as a common signature for MLL-r leukemia[54]. The requirement of these genes for MLL-fusion mediated leukemia have been demonstrated in multiple leukemia models[55, 56]. Additionally, overexpression of HoxA9 was sufficient to transform mouse HSPC and its co-expression with Meis1 accelerated disease development[57, 58].

Therefore, the MLL-fusion-HOXA/MEIS1 axis has become a paradigm in the MLL-r leukemia field.

The heterogeneity of MLL-r leukemia

A unified model of MLL-r leukemogenesis considers all the MLL-r diseases as a single entity, which is encouraging as it suggests there would be common therapeutic targets to ablate all MLL-r leukemia.

However, this model might be an oversimplification given that significant heterogeneity of different subtypes of MLL-r leukemia exists.

MLL-r associates with both ALL and AML, yet the lineage preferences of each type of MLL-fusion are not equivalent. MLL-AF9, -AF10 and -AF6 are associated more frequently with AML than ALL, and

MLL-ELL patients almost exclusively induce AML. In contrast, MLL-ENL is more ALL biased, and

>95% of MLL-AF4 patients develop B-ALL. In MLL-r related T-ALL, ~75% of patients have MLL-ENL or MLL-AF6[35]. In a human HSPC leukemia model, it was demonstrated that compared to MLL-AF9,

MLL-ENL expressing cells have a stronger B-lymphoid propensity[59]. Similarly, it was shown that Mll-

25

Enl but not Mll-Af9 was able to transform T-lineage progenitors in Cre-loxP translocator mice[60]. All these data suggest the MLL-fusions have an instructive role in distinct lineage commitment. Other factors determining the lineage output of the disease include etiology of the transformed cells and microenvironment. The MLL-r infant patients have ALL more frequently than the adult patients, which may be related to the finding that HSC lose lymphoid potential gradually during aging[35].

Experimentally, it has been show the human HSPC from infants transduced with MLL-AF9 can develop

ALL, while only myeloid expansion was obtained when adult HSPC were transformed[61]. A similar result is observed using murine HSPC[62]. Our group has shown that the microenvironmental cues have an important role in the lineage decision of MLL-r leukemia, as exposure to myeloid cytokines can direct

MLL-AF9 leukemic cells towards a myeloid lineage[63]. The different lineage outcome could influence therapeutic decisions, as a recent study showed that in contrast to MLL-r AML, the MLL-r ALL cells were particularly sensitive to proteasome inhibitors[64].

MLL-fusion associated specificity has also been identified even within the disease of the same lineage[65]. A comparison of ALL patients revealed that MLL-AF9 ALL cells do not express CD34, in stark contrast to MLL-AF4 cells[66]. In addition to immunophenotypic differences, a substantial fusion partner-specific transcription and epigenetic program has also been identified[67-69]. Furthermore, the dependency of leukemia cells on particular genes or pathways could vary among different MLL-fusions.

For example, FLT3, which is regarded as a general MLL-fusion target, is required for MLL-ENL but not

MLL-AF9 AML development[70]. The paradigm that HOXA9 is indispensable for MLL-r leukemogenesis is also being challenged. Although the murine MLL-ENL retroviral AML model exhibited a dependence on HoxA9[71] , HoxA9 knockout did not alter AML latency of MLL-AF9 knock- in mice but only the cellular morphology [72]. Additionally, germline loss of HoxA9 did not block the transformation by MLL-GAS7, although AML development was delayed[73]. While these results could be attributed to redundancies from other HOXA genes, several studies have shown that about half of

MLL-r ALL patients do not have upregulation of the whole HOXA loci[67, 74], suggesting that

26 alternative pathways other than HOXA exist for MLL-r leukemia development, especially in ALL.

Strikingly, the species of cells used in a study can also affect the outcome. Runx1 appeared to be a tumor suppressor in a mouse MLL-ENL model, as knockout of Runx1 enhanced the proliferation of AML cells[75]. In contrast, RUNX1 knockdown inhibited proliferation of MLL-AF9 AML derived from human

HSPC as well as cells of MLL-r patient samples. This data suggest that Runx2 may compensate some negative effects caused by loss of Runx1 in mice, however this compensatory mechanism was not engaged in human cells[76].

Overall, all these data indicate that the fusion partners and cell lineages affect the properties of the MLL-r leukemia, arguing that consideration of all MLL-r leukemias as a single entity is not appropriate.

Heterogeneity of MLL-r leukemia castes the caution when we try to extend the findings generated from one MLL-r model to all MLL-r disease. The better way to identify potential therapeutic targets for a certain MLL-fusion is using the accurate model of that fusion, and the human-mouse species difference has to be kept in mind.

Difficulties in MLL-AF4 proB ALL modeling

As we discussed above, with the concern of the heterogeneity of MLL-r leukemia, in order to precisely understand the pathogenesis of MLL-AF4 proB-ALL, the most frequent MLL-r disease, an accurate model is needed. Because of the lack of success using murine cells to model MLL-AF4 proB ALL, people have turned to the human system. However, overexpression of MLL-AF4 in human CD34+HSPC using only resulted in a transient expansion of cells in vitro, and while the engraftment of MLL-

AF4 expressing cells in NSG was higher than the control group, no leukemia developed[77]. The inability to generate a faithful model of MLL-AF4 leukemia has led to the development of alternative hypotheses about t(4;11) pathogenesis, including that an alternative cell of origin other than HSPC is needed, that

MLL-AF4 is unable to transform cells without cooperating oncogenes, or that the reciprocal fusion is the true driver of leukemogenesis.

27

Genetic evidence has shown that MLL translocations occurs in utero[78]. It was reported that bone marrow stromal cells from t(4;11)-pro-B infant ALL patients harbor and express the MLL-AF4 fusion, suggesting that progenitors more primitive than HSC may serve as cell-of-origin for t(4;11)

ALL[79]. To explore this possibility, MLL-AF4 have been introduced into human embryonic stem cells

(hESCs). Enforced MLL-AF4 expression in hESCs accelerated development of hemogenic precursors, however, MLL-AF4 impaired subsequent hematopoietic commitment in favor of an endothelial cell fate, and the transformation of hESC-derived hematopoietic cells was not observed[80]. The short latency of

MLL-AF4 infant B-ALL suggests that secondary mutations are unlikely to be accumulated within this short time period and contribute to leukemia generation. In accordance, recent whole genome sequencing analysis in MLL-r infant ALL samples revealed remarkably few somatic mutations were present. In this study, mutations in genes involved in kinase-PI3K-RAS signaling pathways, including FLT3 and

KRAS, were found in ~2/3 of MLL-AF4 samples, but were mostly subclonal[68]. Other studies accordingly showed that FLT3 mutations are present in less than 20% of MLL-AF4 patients, and only

~25% show RAS mutations[81, 82]. The cooperation between these mutations and MLL-AF4 have been investigated experimentally. KRAS mutation accelerated lymphoid neoplasms progression in MLL-AF4 transgenic mice, through whether the disease was proB ALL requires more detailed characterization[83].

Coexpression of KRAS mutant with MLL-AF4 in human CD34+HSPC failed to immortalize the cells in vitro. In immunodeficient mice, KRAS activation enhanced engraftment and extramedullary hematopoiesis of MA4-expressing HSPCs, yet no leukemia emerged[84]. FLT3 activating mutations have also been expressed in human ESC or HSPC with MLL-AF4, and similarly, activated FLT3 gives some benefit for cell proliferation but does not lead to leukemia[85, 86]. Together, these studies suggest these mutations are unlikely primary tumor drivers. It was reported that the reciprocal fusion protein AF4-

MLL, resulting from the same balanced t(4;11) translocation, but not MLL-AF4, can transform murine

HSPC and induce ALL in mice, arguing that MLL-AF4 might not be the true driver for t(4;11) proB

ALL[87]. This finding is controversial however, since AF4-MLL is not expressed in one third of the t(4;11) patients[88]. Additionally, MLL-AF4 fusion can be formed without rearrangement of t(4;11), but

28 via the insertion of 5’MLL sequence into the AF4 locus, which precludes the genomic recombination of

AF4-MLL[89]. Moreover, MLL-AF4 and AF4-MLL knockdown experiments have shown that t(4;11)- positive cell lines display addiction to MLL-AF4 for cell proliferation and survival, but not to AF4-

MLL[90]. It is possible that both MLL-AF4 and AF4-MLL are oncogenic and contribute to leukemogenesis independently. An examination of the oncogenic role of AF4-MLL in human HSPC would provide useful clarification.

In this study, we found that the AF4 gene has detrimental effects on retroviral titer, which is one of the reasons accounting for the failure of transforming HSPC using the MLL-AF4 . Using MLL-Af4

(5’ of human MLL gene fused to 3’ of murine Af4 gene) to improve retroviral titer, we successfully established a faithful MLL-AF4 proB ALL model that mimics the distinct immunophenotypic and molecular features of the disease. The requirement of the mouse Af4 cDNA to generate high titer virus combined with the necessity to express MLL-Af4 in human hematopoietic stem cells to recapitulate the t(4;11) leukemia phenotype highlights the complexity of accurately modeling human disease in the mouse and has important implications for the development of mouse models of human disease. This MLL-Af4 model can serve as a useful platform to test the function of recently identified cooperating mutations[68] and for the development of customized therapeutic targeting.

2.2 Results

MLL-Af4 produces high titer retrovirus permitting transformation of murine HSPC

We initially sought to establish an MLL-AF4 leukemia model using a retroviral transduction approach.

However, MLL-AF4 retroviral titers were consistently low, as reported by others[40], and thus inadequate for efficient transduction of murine HSPC[91]. This has precluded analyses of the effects of MLL-AF4 on virally transduced murine and human HSPCs. Multiple factors can affect viral titer production, and some human cDNAs have been found to dramatically decrease viral titers[92]. Murine homologs of MLL and its partners have been employed in mouse models[93, 94]. We have observed that retrovirus containing

29 murine Af4, which is highly conserved with human AF4, can be produced at high titers. We therefore constructed an MLL-Af4 retrovirus (Figure 2-2 A). Expression of the MLL-Af4 protein corresponded to that found in t(4;11) leukemia cell lines (Figure 2-2 B). MLL-Af4 retroviral titers were approximately 30- fold higher than with MLL-AF4 (Figure 2-2 C). MLL-Af4 transduction of murine HSPC resulted in immortalization of the cells, not seen using MLL-AF4 (Figure 2-2 D). To assess the leukemogenic potential of MLL-Af4, HSPC were harvested from mice, stimulated with either myeloid or lymphoid cytokines (since cell culture conditions can affect the lineage of the resulting leukemia[95]), transduced, and transplanted into sublethally irradiated C57BL/6 mice. Mice receiving cells transduced with MLL-

Af4 developed acute leukemia with a median latency of 90 days, manifesting similar disease phenotype irrespective of culture conditions (Figure 2-2 E). Leukemia cells with immature myelomonocytic morphology were observed in bone marrow (BM) and peripheral blood (PB) and expressed Gr-1 and

Mac1 but not B220 or CD3, with a subpopulation expressing c-Kit (Figures 2-2 F and G). The MLL-Af4- transplanted mice exhibited significant splenomegaly, and infiltrating leukemic cells were positive for

CD11b and negative for B220 (Figure 2-2 G). Secondary transplant of BM cells confirmed the malignant nature of the disease (Figure 2-2 E). Thus, murine HSPC expressing MLL-Af4 induced AML in vivo and, despite the use of lymphoid conditions, no lymphoid leukemia was observed.

30

31

Figure 2-2. Stably expressed MLL-Af4 in mouse HSPCs induces AML. (A)Schematic of conserved domains contained within the MLL-AF4 and -Af4 fusion proteins. (B) Western blot analysis showed MLL-Af4 expression in transduced Phoenix cells and MLL-AF4 expression in human t(4;11) cell lines (RS4;11 and SEM ). Non-t(4;11) cell lines (U937 and REH) are negative control. anti-MLL antibody detects both wild type N-terminal MLL and fusion proteins. (C) Comparison of retroviral titers of N-terminal MLL, MLL-AF4, and MLL-Af4. Result represents mean +- SD (n=3). (D) Methylcellulose colony-forming assay of mouse HSPCs transduced with N-terminal MLL, MLL- AF4, and MLL-Af4. Results represents mean +- SD (n=3). (E) Kaplan-Meier survival curves of mice transplanted with mouse HSPCs expressing MLL-AF4 or -Af4 using lymphoid (AF4 n=10; Af4 n=10) or myeloid conditions (AF4 n=10; Af4 n=10), and secondary transplantation of MLL-Af4 leukemic cells (n=5). Results were confirmed in 2 independent experiments. (F) Immunophenotype of MLL-Af4 leukemias by flow cytometry. (G) Morphologic and immunohistochemical characterization of MLL-Af4 leukemias showed the immature myelomonocytic leukemic blast cells. Immunohistochemical staining of CD11b and B220 was done on . Scale bar =10 µm (PB and BM) and 50 µm (spleen).

Human CD34+ HSPC expressing MLL-Af4 initiate proB ALL in vivo

Murine genetic models of MLL-AF4 leukemia induce primarily AML or lymphoma, and B-ALL is rarely seen [43, 94, 96]. In contrast, MLL-fusion proteins efficiently induce B-ALL when expressed in human

CD34+ cells and transplanted into immunodeficient mice[59, 63]. To test whether MLL-Af4 induces B-

ALL in a human model, we transduced human CD34+ cells and injected into NOD/SCID/gamma-/- (NSG) mice. As early as 12 weeks post-transplant, the BM and PB of mice showed an expansion of human

CD19+ cells variably expressing CD34 and CD10 (Figure 2-3 A). All animals were leukemic by 22 weeks post-transplant. Splenomegaly was noted and immunohistochemical analysis demonstrated leukemia infiltration into multiple organs (Figures 2-3 B-D). Analysis of PB and BM revealed

CD19+CD33-CD34+ lymphoid blasts that were predominantly CD10 negative, hallmarks of classic t(4;11) proB ALL (Figures 2-3 A and C)[97]. As reported for t(4;11) patients, the lymphoid blasts expressed the myeloid antigen CD15 as well, although the expression of CD65 was less evident compared to patient samples (Figure 2-3 F). Expression of the introduced leukemia oncogene was confirmed in leukemic blasts, and the MLL-Af4 protein was detected at physiological levels comparable to those of MLL-AF4 in a t(4;11) cell line or patient samples (Figures 2-3 G and H).

32

33

Figure 2-3. Human CD34+ cells expressing MLL-Af4 initiate proB ALL in NSG mice. (A) Flow cytometry analysis of BM and PB at indicated time points. (B) Splenomegaly was consistently found in NSG mice reconstituted with MLL-Af4 cells. (C and D) Flow cytometry analysis (C) and paraffin sections (D) of spleen, liver and lung analyzed by hematoxylin and eosin staining (HE) or with human Ki67 showed infiltration of leukemic cells. Scale bar = 50 µm. (E) Wright-Giemsa-stained PB and BM cytospins showed the presence of malignant lymphoid blast cells. Scale bar = 10 µm. (F)Flow cytometry analysis of CD15 and CD65 expression on CD19+ MLL-Af4 leukemic cells and t(4;11) patient xenografts. Two representative experiments are shown. (G) Results of RT-PCR confirmed the expression of MLL-Af4 in leukemic cells. 9 individual mice from 3 independent experiments are shown. Human cells without transduction were used as negative control. (H) Immunoblot analysis showed the expression of MLL-Af4 protein in leukemic cells and MLL-AF4 protein in t(4;11) cell line (RS4;11) and patient samples. Human cells expressing MLL-AF9 were used as control.

Figure 2-4. MLL-Af4 proB-ALL is transplantable. (A) Kaplan-Meier curve of leukemia-free recipient mice. One representative experiment is shown. (B) Flow cytometry analysis showed the disease of secondary leukemic mice had the same proB-ALL phenotype. (C) Compared to disease of primary mice, the CD34+CD10- compartment increased in secondary disease. Two representative experiments are shown. (D) Summary of cell surface marker expression in leukemic mice shown in (C). (E) Summary of MLL-Af4 xenograft experiments.

34

Disease was readily transferred to secondary recipients (Figure 2-4 A) and showed a similar immunophenotype but with a more predominant CD34+CD10- compartment (Figures 2-4 B-D). The disease had high penetrance and the phenotype was reproducible in multiple experiments (Figure 2-4 E).

RNA-Seq analysis of CD19+CD34+ leukemia cells and control human proB cells purified from the BM of NSG mice revealed many MLL-AF4 target genes identified in patient samples were also deregulated in

MLL-Af4 cells[67, 68]. GSEA analysis confirmed significant enrichment of an MLL-AF4 signature

(Figures 2-5 A and B). Thus, expression of MLL-Af4 in human CD34+ cells, from either cord blood or adults, induces proB ALL that mimics the disease found in humans both phenotypically and molecularly.

The ability to faithfully recapitulate t(4;11) proB ALL using transduction of human CD34+ cells contrasts with prior t(4;11) mouse models and demonstrates that the proper combination of species of oncogene and targeting cell can serve as an effective approach to overcome difficulties in disease modeling.

35

Figure 2-5. Gene signatures derived from t(4;11) patients are enriched in MLL-Af4 proB-ALL. (A) GSEA showed that the MLL-AF4 activating/repressing signature derived from two published patient datasets was significantly enriched in MLL-Af4 and control proB cells, respectively. (B) Fold-change ranked heatmap showed significantly differentially expressed MLL-AF4 activating (pink) and repressing (green) signature genes (Stam dataset) between MLL-Af4 leukemic cells and control proB cells. The signature genes were derived by comparing MLL-AF4 patients to non-MLL rearranged ALL patients.

MLL-fusion proteins lead to different developmental stage block of ALL

Though MLL-fusion induced leukemia has been considered as a single entity, heterogeneity of the disease has been observed[66-68], suggesting each type of MLL-fusion protein has its unique characteristics. To test this, we performed a comparative analysis of the B-ALL disease promoted by either MLL-Af4 or

MLL-AF9 using matched units of human CD34+ cells. Immunophenotypically, the MLL-AF9-mediated

B-ALL was distinct, with variable expression of CD10 and little expression of CD34, suggestive of developmental stage differences (Figure 2-6 A). In fact, when compared against defined developmental signatures of human B cell differentiation, the MLL-Af4 leukemia resembled a proB stage while the

MLL-AF9 B-ALL resembled a preB state by transcriptional profiling[98]. Accordingly, developmental stage segregation can be observed when comparing control proB cells with the MLL-AF9 B-ALL but not the MLL-Af4 B-ALL (Figures 2-6 B and C). To validate this finding using a published dataset of clinical samples, we compared t(4;11) and t(9;11) ALL patient samples and observed the same differences in B- cell differentiation stage (Figure 2-6 D). It was reported recently that preB cell receptor (BCR) signaling positive ALL cells rely on different signaling pathways and exhibit distinct kinase inhibitor sensitivity profiles compared to pre-BCR negative ALL cells[99]. We tested for pre-BCR by staining for immunoglobulin u heavy chain (µHC) in our model B-ALL samples and found MLL-AF9 but not MLL-

Af4 B-ALL showed positivity (Figure 2-6 E). BCL6 expression was a surrogate maker of pre-BCR signaling[99], and accordingly, the model MLL-AF9 ALL cells had significantly higher BCL6 compared to MLL-Af4, which was also seen in patient samples (Figure 2-6 F). These findings again demonstrate the unique molecular signals driving these two MLL-fusion leukemias.

36

Figure 2-6. MLL-Af4 and MLL-AF9 lead to block at distinct B cell development stage. (A) Flow cytometry analysis of MLL-Af4 and MLL-AF9 ALL. (B) GSEA result of MLL-Af4 vs -AF9 vs proB showing proB gene signature was enriched in MLL-Af4, and preB gene signature was enriched in MLL-AF9 cells. Less significant enrichment was achieved in MLL-Af4 vs proB comparison. (C and D) Fold-change ranked heatmap showing significantly differentially expressed proB (pink) and preB (green) signature genes between homemade (C) and patient (D) MLL-Af4/AF4 and -AF9 ALL. Patient samples were from Andersson dataset. (E) Flow cytometry staining for surface (s) and cytoplasmic (cy) µHC of MLL-Af4 and -AF9 ALL. Two independent experiments are shown. (F) RNAseq or microarray analysis comparing BCL6 expression of MLL-AF9 cells versus MLL-Af4/AF4 cells in model system and patient samples in Andersson dataset (t(4;11) n=24, t(9;11) n=6) and Stam dataset (t(4;11) n=29, t(9;11) n=8, Probeset=203140_at). The p values were calculated by two-tailed t-test. 37

MLL-Af4 drives a distinct gene expression profile through differential DNA binding

In addition to immunophenotypic differences, heterogeneous gene expression profiles of MLL-fusion

ALL are also observed. A recent study identified the top 100 genes that best discriminate different MLL- fusion B-ALL according to specific translocation partner[68]. Unsupervised hierarchical clustering based on these 100 genes demonstrated the fidelity of these MLL-Af4 and MLL-AF9 model leukemias, with each leukemia associating closely with the respective patient samples (Figures 2-7 A and B). In addition, based on the expression of genes significantly differentially expressed between MLL-Af4 and MLL-AF9

ALL, t(4;11) patient samples clustered closely with MLL-Af4 cells and were readily distinguished from other MLL-fusion samples (Figures 2-7 C and D).

We examined the expression profiles for gene families that distinguish MLL-Af4 and MLL-AF9 B-ALL and identified the HOXA cluster as significantly differentially expressed between these B-ALL samples

(Figure 2-8 A). HOXA9 is considered a bona fide downstream target and one of most critical mediators in

MLL-fusion AML[55]. However, recent analyses showed that approximately half of t(4;11) ALL patients do not have an activated HOXA signature[67, 74]. In our model system, MLL-Af4 ALL does not upregulate HOXA genes compared to control proB cells, in stark contrast to HOXA gene expression in

MLL-AF9 ALL (Figures 2-8 A and B). By comparison, MEIS1, another well-known target of MLL- fusion, is equally expressed in both types of leukemia, and RUNX1, a key target for t(4;11) leukemogenesis[100], is specifically increased only in MLL-Af4 ALL (Figure 2-8 B). Given that our

ALL cells have matched genetic backgrounds, this suggests that the unique fusion protein is the major driving force of the heterogeneous gene expression. Although the DNA binding properties of the different

MLL-fusion proteins remain poorly understood, the presumption in the field is that the fusion proteins are targeted to gene loci via the DNA binding domains of MLL, and thus have similar DNA binding profiles.

However, it has shown that MLL-fusion proteins only bind to a subset of wildtype MLL targets[101], suggesting that the translocation partner genes alter and modify the fusion protein’s binding properties.

To test whether the heterogeneous gene expression is mediated by distinct DNA binding of different

38

MLL-fusions, we performed ChIP-qPCR to compare the chromatin occupancy of MLL-Af4 and MLL-

AF9 at these specific loci. Indeed, chromatin occupancy correlated with gene expression, with no binding of MLL-Af4 at the HOXA locus nor MLL-AF9 binding at the RUNX1 locus (Figure 2-8 C). Several additional differentially-expressed MLL-Af4 and -AF9 genes were also found to correlate with specific

DNA binding of the different MLL-fusion proteins (Figure 2-8 D). These results demonstrate that differential DNA binding of MLL-fusion proteins contributes to distinct gene expression profiles.

39

Figure 2-7. MLL-Af4 gene expression profile recapitulates t(4;11) specific molecular signature. (A and B) Principal component analysis (A) and unsupervised hierarchical clustering (B) of homemade MLL-Af4 and -AF9 leukemia together with MLL-fusion ALL patient samples based on the expression of a 100-gene discriminator from Andersson et al. (C and D) The same analysis was performed based on the expression of 430 genes which were most significantly differential expressed between homemade MLL-Af4 and -AF9 leukemic cells (Appendix Table 5).

Figure 2-8. MLL-fusions promote distinct gene expression profiles via differential DNA binding. (A) Heatmap of HOXA gene expression of MLL-AF9, MLL-Af4 and control proB cells. (B) Differential activation of reported MLL-fusion targets between MLL-AF9 and MLL-Af4 cells compared to proB cells. Expression data was derived from RNAseq, normalized by mean=0, variance=1. (C) ChIP-qPCR analysis showed that MLL-Af4 and -AF9 have distinct chromatin occupancy at target gene loci correlating with gene expression in (B). (D and E) Heatmap (D) and ChIP-qPCR analysis (E) of more selected genes specifically expressed in MLL-Af4 or -AF9 ALL showing corresponding chromatin occupancy of MLL-fusions at their specific target gene loci. For ChIP-qPCR, IVL loci was used as negative control. The result represents mean and SD, n=3 biological replicates.

40

MLL-Af4 myeloid cells retain lymphoid lineage potential

Although the tight association of t(4;11) with B-cell ALL has been apparent since the translocation was first identified, it has been impossible to determine whether this was due to intrinsic MLL-AF4 signaling differences or was a result of the translocation occurring in a particular cell type biased towards a B-cell fate. We and others have previously shown that human CD34+ cells expressing MLL-AF9 generate AML in immunodeficient mice after priming in myeloid culture conditions[59, 63, 102]. We established such myeloid cell cultures with MLL-Af4. Surprisingly, these MLL-Af4 cells retained a population of

CD19+CD33- cells not present in paired MLL-AF9 cultures (Figure 2-9 A). When cells were transferred to B-lymphoid promoting conditions, CD19+CD33- cells rapidly expanded in MLL-Af4 but not MLL-

AF9 cultures (Figure 2-9 B). Similarly, upon injection into NSG mice, the MLL-Af4 myeloid cells reproducibly induced human B-ALL with a CD34+CD10- proB cell phenotype, while the MLL-AF9 cells invariably resulted in CD33+CD19- AML (Figures 2-9 C and D). B-ALL also developed in vitro and in vivo from a MLL-Af4 myeloid clone having no detectable CD19+ cells (Figures 2-9 E and F).

41

Figure 2-9. MLL-Af4 maintains lymphoid potential after myeloid priming. (A) Percentage of CD33-CD19+ cells in MLL-Af4 and -AF9 cultures primed in myeloid conditions for 5 weeks. Five independent experiments are shown. (B) CD33 and CD19 expression of MLL-Af4 and -AF9 cells in week 2 B-cell culture condition after initial priming in myeloid condition for 5 weeks. (C) Flow cytometry of leukemia initiated by MLL-Af4 or -AF9 myeloid- primed cells in NSG mice. (D) Leukemia initiated by myeloid-primed MLL-Af4 cells had proB immunophenotype. Two independent experiments are shown. (E and F) A particular MLL-Af4 myeloid culture without detectable CD33-CD19+ cells gave rise to CD19+ B cells both in-vitro (E) and in NSG mice (F).

Figure 2-10. MLL-Af4 phenotypic myeloid cells keep an active lymphoid program. (A) Flow cytometry analysis and Wright-Giemsa staining of CD33+CD19- sorted MLL-Af4 cells before and after lymphoid culture switch. Scale bar = 10 µm. (B) Heatmap showed increased expression of lymphoid genes and decreased expression of myeloid genes in CD33+CD19- cells expressing MLL-Af4 compared to those expressing MLL-AF9. All genes shown achieved a significance of p≤0.05 with fold change≥1.5. (C) Dataset enrichment analysis by LR-Path showed B cell signature is associated with CD33+CD19- cells expressing MLL-Af4 cells compared to those expressing MLL-AF9. Datasets enriched in genes upregulated or downregulated by MLL-Af4 were colored red and green, respectively.

42

Moreover, sorted CD33+CD19- MLL-Af4 cells reproducibly generated CD33-CD19+ B-lymphoid cells under B-cell growth conditions (Figure 2-10 A), indicating the CD33+CD19- MLL-Af4 cells harbor latent B-cell potential refractory to myeloid priming. In contrast to data showing MLL leukemia lineage is ontogenically determined, with AML resulting from adult-origin and ALL from fetal-origin HSPC[61],

MLL-Af4 cells derived from adult CD34+ cells showed identical lymphoid persistence as CD34+ cells obtained from cord blood (data not shown). These results suggest the strong lymphoid preference is an

MLL-Af4 intrinsic property that overrides the lineage instruction from microenvironment and cell-of- origin influence. RNA-Seq analysis of CD33+CD19- cells sorted from four sets of genetically matched

MLL-Af4/MLL-AF9 myeloid cultures showed uniform upregulation of key B-lymphoid genes and a decrease in specific myeloid genes in MLL-Af4 cells relative to matched MLL-AF9 cells, demonstrating that MLL-Af4 maintains an active lymphoid program responsible for the B-cell bias of human cells

(Figures 2-10 B and C). t(4;11) ALL acquires resistance to CD19-targeted therapy by myeloid differentiation

A recent treatment advance in relapsed B-ALL employs a bispecific antibody, blinatumomab, that specifically targets CD19 on B cells[103]. In a pediatric patient with refractory t(4:11) proB ALL treated with blinatumomab, t(4;11) AML relapse was observed (Figures 2-11 A and B)[104]. These t(4;11) myeloid cells promoted B-ALL in NSG mice, even when using sorted CD33+CD19- cells (Figure 2-11

C). A marker chromosome indicated the B-ALL cells that expanded in the mouse were clonally related to the initial myelomonocytic cells (data not shown). QPCR analysis of this AML in comparison to standard

AML samples showed significant upregulation of some of the same B-cell genes identified in the MLL-

Af4 cells (Figure 2-11 D). Thus similar to the effects seen with the MLL-Af4 myeloid cultures, a persistent lymphoid program resistant to environmental induction and resulting in potent B-lymphoid preference is associated with the t(4;11) fusion protein, showing the utility of our MLL-Af4 model system as applied to clinical disease. We also identified an adult patient with t(4;11) proB ALL treated with blinatumomab who relapsed with morphologic, cytochemical, and immunophenotypic evidence of

43 differentiation towards the monocytic lineage (Figures 2-11 E-G). This phenotypic flexibility represents a novel escape mechanism from CD19-targeted treatment for patients with t(4:11) ALL, and has also been reported recently in t(4;11) patients receiving chimeric antigen receptor T-cell (CAR-T) therapies directed against CD19[105]. Therapeutic strategies may need to be customized for this poor prognosis leukemia showing phenotypic plasticity with transcriptional lymphoid persistence under selective pressure of

CD19-directed therapy. Our model accurately recapitulates both a de novo ALL stage and a refractory

‘AML’ stage, enabling studies to proceed that may reveal new insights into molecular mechanisms and permit development of novel therapies.

2.3 Discussion

We and others have previously observed very low retroviral titers from producer cells expressing MLL-

AF4. [40, 77]. Gene dosage is one critical factor influencing the transformation capacity of oncogenes[106]. A retroviral approach entailing variable titers and transduction efficiencies cannot efficiently achieve gene dosage thresholds required for transformation. Therefore, the detrimental effects on viral titer caused by AF4 gene could be one reason that explains why previous attempts using MLL-

AF4 to transform mouse and human HSPC were not successful. Although the mouse and human AF4 proteins are highly conserved, the level of conservation in the cDNA sequences is lower and we have observed distinct species-related differences between the viral titer achieved using mouse Af4 and human

AF4 genes. Multiple factors can affect viral titer production including inhibitory sequences that affect nucleic acid sequence-dependent steps in virus production, inappropriate signals present in some sequences that inhibit virus production, potential toxic effects of the gene sequence on producer cells, and inappropriate splicing due to aberrant use of alternative splice acceptor/donor sequences in the insert. As shown in a recent library screen, some human cDNAs can dramatically decrease viral titer, [92]. Inspired by this observation, we fused human MLL to murine Af4. Strikingly, expression of MLL-Af4 in human

CD34+ cells faithfully recapitulates the proB ALL observed in patients with the t(4:11). This is highlighted by both the immunophenotypic analysis and the gene expression signature identified by

44

RNAseq. These results demonstrate that, at least for MLL-Af4 driven B-ALL, prehematopoietic mesodermal or hemangioblast precursors are not required as initiating cells and the reciprocal AF4-MLL fusion is dispensable. Whether these observations are reflected in pathogenesis of t(4;11) patients remains to be determined.

45

Figure 2-11. Phenotypic flexibility of t(4;11) ALL contributes to resistance to CD19-taregeted therapy. (A) Schematic of lineage progression of the pediatric t(4;11) patient sample. (B) Wright-Giemsa-staining of BM of pediatric t(4;11) patient with relapse AML. Scale bar=10 µm. (C) CD33/CD19 expression of leukemia cells from NSG mice reconstituted with the pediatric t(4;11) relapse AML sample. (D) qPCR results of selected lineage genes in CD33+CD19- sorted cells for t(4;11) AML and cytogenetically normal AML. n=3 technical replicates, error bars represent SD. (E) Immunophenotype comparison of the adult t(4;11) patient sample before and after blinatumomab treatment. (F) Wright-Giemsa-staining of BM of the adult t(4;11) patient before and after blinatumomab treatment. Scale bar=10 µm. (G) Alpha-naphthyl-butyrate and CD33 staining of the adult t(4;11) patient sample relapsing from blinatumomab treatment. Scale bar= 50 µm.

We have observed distinct lineage preferences of MLL-Af4-transduced mouse and human HSPCs. Mouse models of other oncogenes associated with human ALL have also resulted in the unexpected generation of myeloid leukemia. A myeloid lineage preference in murine models has previously been observed for

E2A-PBX1, a fusion oncoprotein associated exclusively with human preB cell ALL[107]. The factors that mediate myeloid lineage preference of E2A-PBX1 and MLL-Af4 in murine cells remain unknown.

However, the distinct linage output associated with the human cell model highlights the limitation of using mouse cells for modeling human disease.

Many common fusion partners of MLL translocation have been identified as components of SEC[36] and are involved in complex interactions with the H3K79 histone methyltransferase DOT1L [40]. Thus, the aberrant transcriptional elongation and H3K79 methylation that lead to dysregulation of target genes are considered the general mechanism of MLL-fusion mediated leukemogenesis. Indeed, a shared molecular signature has been identified for MLL-fusion leukemia irrespective of fusion partner or lineage[54], leading to the notion that all MLL-fusion proteins work in a similar fashion by dysregulating the same pathways, the paradigm of which is activation of HOXA and MEIS1 genes. However, biological differences observed between subtypes of MLL-fusion leukemia raise questions regarding this view. 50% of t(4;11) ALL patients do not show HOXA gene activation, and low HOXA gene expression is actually associated with a worse prognosis[67, 74]. Accordingly, our MLL-Af4 ALL cells do not show increased

HOXA expression compared to normal proB cells, indicating that at least for a set of MLL-fusion ALL,

46 activation of HOXA is not required for leukemogenesis. Therefore, since most of our knowledge about

MLL-fusion leukemia is generated from mouse models of AML, primarily MLL-AF9 and MLL-ENL, care must be taken in generalizing these findings to all MLL-fusion disease.

The instructional role of the fusion partner to influence the phenotype of the disease has been reported previously, primarily focusing on lineage preference[59, 60, 63]. Here we show that even within the same lineage, lymphoid in this case, different MLL-fusion partner proteins can instruct the disease to distinct stages of differentiation. In contrast to MLL-Af4 ALL cells, which faithfully manifest a proB immunophenotype, MLL-AF9 ALL cells resembled a later preB stage. Additionally, MLL-AF9 cells express both surface and cytoplasmic uHC, suggesting they have positive pre-BCR signaling. A recent report showed that pre-BCR+ ALL is a distinct subtype from pre-BCR(-) ALL, relying on a different signaling pathway and showing a selective sensitivity to pre-BCR tyrosine kinase inhibitors[99]. These data suggest MLL-AF4 and MLL-AF9 leukemia may depend on diverse signaling pathways and that therapeutic targets identified in one MLL subtype may not be applicable to others. As each MLL-fusion leukemia could have its own Achilles' heel, customized therapy may need to be introduced that is specific to each.

Although a common signature was identified for all MLL-fusion leukemia, significant fusion partner transcriptome and DNA methylome heterogeneity has been found[67-69]. The mechanisms accounting for this heterogeneity are not fully understood, and due to limited numbers of leukemia cells and complex genetic background, it is difficult to gain insight from patient samples. We show here that matched MLL-

Af4 and MLL-AF9 ALL cells with the same genetic background can recapitulate the fusion partner specific gene signature derived from patient samples. In addition, the signature generated from our model leukemia can be utilized to classify patient samples, suggesting the fusion partner protein is one of the major forces for gene expression diversity. Importantly, we demonstrate differential DNA binding of the fusion protein is one molecular mechanism for differential gene regulation. This is the first evidence showing the DNA binding capacity of MLL-fusion proteins is not completely attributable to the MLL

47 portion, implicating the fusion partners in this function. Additionally, previous studies have shown that different MLL-fusions could recruit distinct protein partners[36], which could build another layer of regulation causing gene expression heterogeneity beyond DNA binding. All of these could also be critical mechanisms for the instructional role of MLL-fusions leading to different lineage or developmental stage outcomes.

In patients, MLL-AF4 is almost exclusively associated with B-ALL, while MLL-AF9 is more prevalent in AML[35]. In line with the instructional role of the fusion partner, this different lineage preference is also recapitulated in our comparative study, reflected by the fact that MLL-Af4 cells are resistant to environmental myeloid redirection and maintain a persistent active lymphoid program. While MLL-AF9 cells acquire a stable myeloid fate after myeloid priming and give rise to AML, MLL-Af4 cells return to a lymphoid fate once the environmental pressure is released. Strikingly, this assumed myeloid status appears to be a novel mechanism for t(4;11) disease to escape from blinatumomab therapy, and for the development of resistance to other CD19-directed immunotherapies, such as CAR-T therapy[105].

Clinically, how these myeloid cells with lymphoid potential respond to traditional AML therapy compared to conventional AML cells is unknown. However, our MLL-Af4 myeloid cultures will be useful to evaluate the drug sensitivity for eliminating resistant disease.

In summary, our results and correlating observations in patients demonstrate that MLL-fusion disease is not a single genetic entity. Although different MLL-fusion proteins share some common properties, each has its own genetic and biological features associated with particular fusion partner proteins. These differences could potentially impact response to therapy. Our MLL-Af4 model will be a valuable tool to study this most prevalent and poor prognosis MLL-fusion leukemia.

48

2.4 Experimental Procedures

Cells and culture

Patient samples were acquired following informed consent in accordance with the Declaration of Helsinki and under protocols approved by the institutional review board. Human umbilical cord blood cells (CB) or adult PB progenitor cells (PBPC) were obtained by the Translational Trials Support Laboratory at

CCHMC under an approved protocol. CD34+ cells were enriched using CD34+ selection kit (Miltenyi).

To prime cells into myeloid lineage, MLL-Af4 and -AF9 cells were cultured in IMDM with 10% fetal bovine serum (FBS) and supplemented with SCF (Stem cell factor), IL-3, IL-6, FLT3L and TPO

(Thrombopoietin) (all at 10 ng/mL). For B cell assay, myeloid-primed cells were transferred on to MS-5 stroma cells, maintained in aMEM with 10% FBS and supplemented with SCF, FLT3L, and IL-7 (10 ng/mL). Half of the media/cells were removed and replaced with fresh media weekly.

Retroviral production

AF4, Af4 and AF9 cDNAs were ligated to 5’ MLL and cloned in the MSCV retroviral vector. An in- frame FLAG tag was inserted between MLL residue 1404 (in 11) and the beginning of the partner protein sequence. Retrovirus for transduction of murine cells was packaged in Phoenix cells as described[108]. Viral titers were determined by infection of Rat1A cells with Phoenix retroviral supernatants followed by selection in G418. Virus for human cell transduction was produced using 293T cells transfected with MSCV-MLL-fusion vectors, together with envelope RD114 and the gag-pol M57 constructs in 10 cm plates. 24 h after transfection, the retroviral supernatant were collected every 12h for

3 collections.

Flow cytometry and Cell Sorting

Cells from mouse tissues were incubated with nonspecific binding blocker (anti-mouse/human

CD16/CD32 Fc g receptor; BD) before staining. (all BD unless noted) used were PE-CD10

49

(HI10a), APC- and PE-CD33 (WM53), PECy5-CD34 (581), V450-human CD45 (HI30), APCCy7-mouse

CD45 (30-F11), PE-CD13 (WM15), PECy7-CD19 (5J25C1), PE-CD65(VIM8), FITC-CD15 (MMA),

VioBlue-CD19 (6D5, Miltenyi), FITC-and BV421- uHC (H15101 Caltag Laboratories; MHM88,

Biolegend), FITC- c-Kit and PE-CD11b, Gr-1, CD3, or B220 antibodies (eBioscience). Intracellular staining was performed using Cytofix/Cytoperm kit (BD). Cells were analyzed on FACSCanto flow cytometer (BD) or sorted on FACSAria (BD) or MoFlo XDP (Beckman Coulter), and the data was analyzed with FloJo software (TreeStar).

Methylcellulose colony-forming assays

Infection of lineage-depleted (Lin-) BM cells obtained from C57BL6 mice five days after 5-fluorouracil treatment and culture of the transduced progenitor cells in methylcellulose were performed as previously described[108]. 30,000 cells were used for each transduction. The BM progenitor cells were centrifuged in retroviral supernatant at 2500 g for 4 hours at 33 oC (spinoculation). The cells were incubated in fresh media with appropriate growth factors for 20 hours and spinoculation was repeated. Transformation capability of MLL-fusion constructs were examined in duplicate in at least 2 independent methylcellulose colony-forming assays.

Mouse transplantation

Transplantation was performed using both myeloid and lymphoid conditions. For the myeloid reconstitution assays, 6 week old C57/BL6 mice were pretreated with 5- fluorouracil at 150 mg/kg by intravenous injection and BM cells were harvested 5 days later. Lin- cells were selected using columns

(Miltenyi) and cultured in RPMI media containing β-mercaptoethanol 0.05 mM, 10% FBS supplemented with 100 ng/ml SCF, 10 ng/ml IL-3, and 10 ng/ml IL-6 (R&D Systems, Minneapolis, MN). For the lymphoid conditions, Lin- BM cells were harvested from 6 week old C57/BL6 mice without 5- fluorouracil pretreatment. Lin- BM cells were cultured in 100 ng/ml SCF, 10 ng/ml IL-7, 10 ng/ml IL-6, and Flt-3 ligand 10 ng/ml (R&D Systems)[95]. The BM progenitor cells were centrifuged in retroviral

50 supernatant at 2500 g for 4 hours at 33 oC. For each mouse, 30,000 Lin- cells were spinoculated separately, and the transduced cells were not pooled prior to transplant. The cells were incubated in fresh media with appropriate growth factors for 20 hours and spinoculation was repeated. Reconstitution of sub-lethally irradiated C57BL/6 mice with transduced progenitors was performed as described previously[108]. Each mouse was transplanted by intravenous injection with transduced Lin- BM cells.

For histological analysis, tissues were fixed in formalin, sectioned, and stained with hematoxylin and eosin. For immunohistochemical analysis, tissues were snap frozen in Tissue-Tek O.C.T. compound

(Sikura), sectioned, and stained with antibodies to CD11b and B220. No mice were excluded from analysis. All experiments were performed in accordance with U of Chicago institutional guidelines.

Xenograft transplantation

500×105 CD34+cells were used for each transduction. CD34+ cells were pre-stimulated in IMDM with

10%FBS, SCF, FLT3L, and THPO (100 ng/mL) for 24 hours. Retronectin-coated plates were preloaded three times with 3 mL retroviral supernatant by centrifuging at 2200 rpm and 10 °C for 25 min.

Stimulated cells were cultured in the presence of 3 mL retroviral supernatant on virus-loaded plates

(Takara) overnight, 3mL fresh retroviral supernatant was replaced, and cells were cultured for another 6 hours. To induce acute leukemia in NSG mice, 6- to 12-week-old mice were conditioned with 30 mg/kg busulfan (Sigma) through intraperitoneal injection 24 hours before transplantation, 100-150×105 MLL-

Af4 or -AF9 transduced cells were transplanted through intrafemoral injection immediately after transduction. For cell line experiments, MLL-Af4 or -AF9 cells were primed in myeloid conditions for 4-

8 weeks, and then 0.5-1×106 cells were injected through tail vein. Mice were sacrificed when signs of illness were observed. Organs were homogenized and processed for flow cytometry or fixed in 10% formalin for histopathologic analysis. In serial transplantation, 1×106 BM cells were injected through tail vein. For patient sample study, 1×106 t(4;11) AML cells were injected through tail vein. Each experiment was performed at least in three independent replicates. ProB ALL was successfully induced in all 4

51 independent MLL-Af4 transductions. All experiments were performed in accordance with CCHMC institutional guidelines.

Western blotting

Nuclear lysates were obtained using NE-PER nuclear extraction kit (Thermo Scientific). Protein samples were run on 6% or 4-15% polyacrylamide gels and transferred overnight at 4 °C to a nitrocellulose membrane. The primary antibodies used were anti-Flag (Cell Signalling Technology #2368 and Sigma

M2), anti-Lamin B2 (Cell Signalling Technology #12255), anti-MLL (Bethyl A300-086A) and anti-

(NeoMarker ACTN05). The secondary antibodies were HRP-linked goat anti-rabbit IgG and anti-mouse

IgG at 1:1000 (Cell Signaling Technology #7074, #7076), the signal of which was developed through

ECL reaction; or goat anti-rabbit IRDye 800RD and goat anti-mouse IRDye 680RD (Odyssey) at

1:10000, where the signal was visualized by fluorescent illumination (Odyssey CLx).

Chromatin immunoprecipitation (ChIP)

MLL-Af4 and -AF9 ALL cells were harvested from mice BM and spleen and subjected for ChIP. The

ChIP assay was performed as described previously [109](more details in Chapter 3). Dynabeads protein

G (Invitrogen) pre–incubated with BSA and antibody against Flag (Sigma, M2) were used for IP. The immunoprecipitated DNA was purified using Agencourt AMPure magnetic beads (Beckman Coulter) according to the manufacturer’s instructions, and analyzed by qPCR using SYBR Green technology

(Applied Biosystems). The chromatin enrichment of each gene locus was calculated by standard curve method, normalized to 1% input. Relative enrichment values were normalized to a negative control region of the genome (IVL gene promoter). Primers are listed in Appendix Table 1.

RNA isolation, RT-PCR and RNA sequencing

Human CD45+ CD19+ cells were sorted from BM of 6 individual MLL-Af4 (3 PBPC based and 3 CB based, generated by 2 independent transduction) and 3 individual MLL-AF9 (CB based, generated by one transductions) leukemic mice. For control proB cells, non-transduced CD34+ CB cells were transplanted

52 into NSG mice, 8 weeks after transplantation human CD45+CD19+CD34+ proB cells were sorted from

BM of 3 mice. For myeloid-primed culture study, CD33+CD19- cells were sorted from 4 pairs of MLL-

Af4 and -AF9 clones (2 PBPC based and 2 CB based, generated by 4 independent transductions) that had been cultured in myeloid conditions for 6-8 weeks. Patient samples were similarly sorted. Total RNA was isolated from sorted cells using RNeasy Mini Kit (Qiagen).

For RT-PCR, RNA was reversed transcribed using MuLV Reverse Transcriptase and random hexamers

(Applied Biosystems). The cDNA was then subject to qPCR using SYBR Green technology (Roche).

Expression level was calculated by ∆∆Ct method, normalized to GAPDH. Primers are listed in Appendix

Table 1.

For RNA sequencing, the integrity of RNA was analyzed by Bioanalyzer (Agilent). RNA from each individual mouse was processed separately without pooling.1 µg total RNA was used for poly(A) RNA selection, followed by cDNA synthesis using PrepX mRNA Library kit (WaferGen) and Apollo 324 NGS automatic library prep system. Sample-specific index was added to the adaptor-ligated cDNA by PCR with index-specific primers for 13 cycles. The cluster generation of indexed libraries was carried out on cBot system (Illumina) using Illumina’s TruSeq SR Cluster kit v3, and then sequenced on Illumina HiSeq system using TruSeq SBS kit to generate single-end 50 cycle reads. 20-50 million reads were generated for each sample.

Analysis of RNA sequencing data

Sequence reads were aligned to the human reference genome using the TopHat aligner, and reads aligning to each known transcript were counted using Bioconductor packages for next-generation sequencing data analysis[110]. Transcript expression levels were estimated as reads per kilobase of transcript per million mapped reads (RPKM). For lymphoid leukemia study, RPKM data were imported into Qlucore Omics

Explorer 3.1 software (Qlucore) for further analysis. To identify differentially expressed genes between

MLL-Af4 and MLL-AF9 myeloid cells, the analysis was performed in edgeR Bioconductor package

53 using generalized linear model likelihood ratio test for paired samples[111], with a cutoff of FDR≤0.1 and fold change ≥1.5.

Pathway enrichment analysis

To obtain MLL-AF4 patient gene signature, expression data from Stam et al. were downloaded from

GEO database (GSE19475), expression data from Andersson et al. were obtained from Dr.

Andersson[68], and then imported into Qlucore. By comparing MA4 patients to non-MLL-fusion patients, significant differential genes were selected by built-in statistical functions (p≤0.05, FDR≤0.1, fold change ≥2) and defined as MA4 signature (Appendix Table 2). The expression of MA4 signature genes was evaluated in MLL-Af4 leukemic cells compared to control proB cells, the significant differentially expressed genes between two groups (p≤0.05) were ranked according to fold change as a heatmap and colored depending on whether they are upregulated (pink) or downregulated (green) in MA4 patients. GSEA analysis was performed as described [112]. The same approach was used to evaluate B cell developmental stage specific genes association with MLL-Af4 and -AF9 leukemic cells. The proB and preB signatures were derived from Hystad et al [98]( Appendix Table 3). For myeloid cell studies, the pathway enrichment analysis was performed separately for upregulated and downregulated genes using the LRpath methodology with the gene lists from the MSigDB database[113].

Cluster analysis

In order to test the similarity of MLL-Af4 gene expression to those of MA4 patients relative to patients with other MLL-fusions, the RPKM data of our MLL-Af4, -AF9 and control proB samples plus the

Andersson dataset were imported into Qlucore. The batch effect of different datasets was corrected automatically by using “Eliminated Factor” function of the software. After correction, built-in principal component analysis and unsupervised hierarchical clustering were performed based on the expression of a 100-gene discriminator derived by Andersson et al (Appendix Table 4), which best associates patient samples according to MLL-fusion partners. In addition, to test whether the gene signature derived from

54 our model can be used to discriminate patient samples, the significant differentially expressed genes

(p≤0.05, FDR≤0.1, fold change ≥1.5) between MLL-Af4 and MLL-AF9 CD45+ CD19+ leukemic cells were determined. To avoid batch effects that could skew the data, the comparison was performed using genes not having consistent significant variation between the two datasets irrespective of MLL fusions

(~70% of total genes). A list of 430 genes was generated for cluster analysis (Appendix Table 5).

2.5 References

1. Yokoyama, A., et al., Leukemia proto-oncoprotein MLL is proteolytically processed into 2 fragments with opposite transcriptional properties. Blood, 2002. 100(10): p. 3710-8. 2. Yokoyama, A., et al., Leukemia proto-oncoprotein MLL forms a SET1-like histone methyltransferase complex with menin to regulate Hox gene expression. Mol Cell Biol, 2004. 24(13): p. 5639-49. 3. Milne, T.A., et al., MLL targets SET domain methyltransferase activity to Hox gene promoters. Mol Cell, 2002. 10(5): p. 1107-17. 4. Hsieh, J.J., et al., Proteolytic cleavage of MLL generates a complex of N- and C-terminal fragments that confers protein stability and subnuclear localization. Mol Cell Biol, 2003. 23(1): p. 186-94. 5. Yokoyama, A., et al., MLL becomes functional through intra-molecular interaction not by proteolytic processing. PLoS One, 2013. 8(9): p. e73649. 6. Katsani, K.R., et al., A homeotic mutation in the trithorax SET domain impedes histone binding. Genes Dev, 2001. 15(17): p. 2197-202. 7. Jude, C.D., et al., Unique and independent roles for MLL in adult hematopoietic stem cells and progenitors. Cell Stem Cell, 2007. 1(3): p. 324-37. 8. McMahon, K.A., et al., Mll has a critical role in fetal and adult hematopoietic stem cell self- renewal. Cell Stem Cell, 2007. 1(3): p. 338-45. 9. Yu, B.D., et al., Altered Hox expression and segmental identity in Mll-mutant mice. Nature, 1995. 378(6556): p. 505-8. 10. Yagi, H., et al., Growth disturbance in fetal liver hematopoiesis of Mll-mutant mice. Blood, 1998. 92(1): p. 108-17. 11. Gan, T., et al., Developmentally induced Mll1 loss reveals defects in postnatal . Leukemia, 2010. 24(10): p. 1732-41. 12. Ernst, P., et al., An Mll-dependent Hox program drives hematopoietic progenitor expansion. Curr Biol, 2004. 14(22): p. 2063-9. 13. Artinger, E.L., et al., An MLL-dependent network sustains hematopoiesis. Proc Natl Acad Sci U S A, 2013. 110(29): p. 12000-5. 14. Terranova, R., et al., Histone and DNA methylation defects at Hox genes in mice expressing a SET domain-truncated form of Mll. Proc Natl Acad Sci U S A, 2006. 103(17): p. 6629-34. 15. Mishra, B.P., et al., The histone methyltransferase activity of MLL1 is dispensable for hematopoiesis and leukemogenesis. Cell Rep, 2014. 7(4): p. 1239-47. 16. Yokoyama, A., et al., The menin tumor suppressor protein is an essential oncogenic cofactor for MLL-associated leukemogenesis. Cell, 2005. 123(2): p. 207-18. 17. Yokoyama, A. and M.L. Cleary, Menin critically links MLL proteins with LEDGF on cancer- associated target genes. Cancer Cell, 2008. 14(1): p. 36-46.

55

18. Okuda, H., et al., MLL fusion proteins link transcriptional coactivators to previously active CpG- rich promoters. Nucleic Acids Res, 2014. 42(7): p. 4241-56. 19. Eidahl, J.O., et al., Structural basis for high-affinity binding of LEDGF PWWP to mononucleosomes. Nucleic Acids Res, 2013. 41(6): p. 3924-36. 20. Barski, A., et al., High-resolution profiling of histone methylations in the . Cell, 2007. 129(4): p. 823-37. 21. Milne, T.A., et al., MLL associates specifically with a subset of transcriptionally active target genes. Proc Natl Acad Sci U S A, 2005. 102(41): p. 14765-70. 22. Li, B.E., et al., Distinct pathways regulated by menin and by MLL1 in hematopoietic stem cells and developing B cells. Blood, 2013. 122(12): p. 2039-46. 23. Caslini, C., et al., The amino terminus targets the mixed lineage leukemia (MLL) protein to the nucleolus, nuclear matrix and mitotic chromosomal scaffolds. Leukemia, 2000. 14(11): p. 1898- 908. 24. Xia, Z.B., et al., MLL repression domain interacts with histone deacetylases, the polycomb group proteins HPC2 and BMI-1, and the corepressor C-terminal-binding protein. Proc Natl Acad Sci U S A, 2003. 100(14): p. 8342-7. 25. Allen, M.D., et al., Solution structure of the nonmethyl-CpG-binding CXXC domain of the leukaemia-associated MLL histone methyltransferase. EMBO J, 2006. 25(19): p. 4503-12. 26. Muntean, A.G., et al., The PAF complex synergizes with MLL fusion proteins at HOX loci to promote leukemogenesis. Cancer Cell, 2010. 17(6): p. 609-21. 27. Rosonina, E. and J.L. Manley, From transcription to mRNA: PAF provides a new path. Mol Cell, 2005. 20(2): p. 167-8. 28. Milne, T.A., et al., Multiple interactions recruit MLL1 and MLL1 fusion proteins to the HOXA9 locus in leukemogenesis. Mol Cell, 2010. 38(6): p. 853-63. 29. Krivtsov, A.V. and S.A. Armstrong, MLL translocations, histone modifications and leukaemia stem- cell development. Nat Rev Cancer, 2007. 7(11): p. 823-33. 30. Mohan, M., et al., Licensed to elongate: a molecular mechanism for MLL-based leukaemogenesis. Nature reviews. Cancer, 2010. 10: p. 721-728. 31. Caslini, C., et al., Interaction of MLL amino terminal sequences with menin is required for transformation. Cancer Res, 2007. 67(15): p. 7275-83. 32. Mereau, H., et al., Impairing MLL-fusion gene-mediated transformation by dissecting critical interactions with the lens epithelium-derived growth factor (LEDGF/p75). Leukemia, 2013. 27(6): p. 1245-53. 33. Ayton, P.M., E.H. Chen, and M.L. Cleary, Binding to nonmethylated CpG DNA is essential for target recognition, transactivation, and myeloid transformation by an MLL oncoprotein. Mol Cell Biol, 2004. 24(23): p. 10470-8. 34. Muntean, A.G., et al., MLL fusion protein-driven AML is selectively inhibited by targeted disruption of the MLL-PAFc interaction. Blood, 2013. 122(11): p. 1914-22. 35. Meyer, C., et al., The MLL recombinome of acute leukemias in 2013. Leukemia, 2013. 27: p. 2165- 76. 36. Lin, C., et al., AFF4, a component of the ELL/P-TEFb elongation complex and a shared subunit of MLL chimeras, can link transcription elongation to leukemia. Mol Cell, 2010. 37(3): p. 429-37. 37. Taki, T., et al., AF5q31, a newly identified AF4-related gene, is fused to MLL in infant acute lymphoblastic leukemia with ins(5;11)(q31;q13q23). Proc Natl Acad Sci U S A, 1999. 96(25): p. 14535-40. 38. Shilatifard, A., et al., An RNA polymerase II elongation factor encoded by the human ELL gene. Science, 1996. 271(5257): p. 1873-6. 39. Peterlin, B.M. and D.H. Price, Controlling the elongation phase of transcription with P-TEFb. Mol

56

Cell, 2006. 23(3): p. 297-305. 40. Yokoyama, A., et al., A Higher-Order Complex Containing AF4 and ENL Family Proteins with P- TEFb Facilitates Oncogenic and Physiologic MLL-Dependent Transcription. Cancer Cell, 2010. 17: p. 198-212. 41. Mohan, M., et al., Linking H3K79 trimethylation to Wnt signaling through a novel Dot1- containing complex (DotCom). Genes Dev, 2010. 24(6): p. 574-89. 42. Schubeler, D., et al., The histone modification pattern of active genes revealed through genome- wide chromatin analysis of a higher eukaryote. Genes Dev, 2004. 18(11): p. 1263-71. 43. Krivtsov, A.V., et al., H3K79 Methylation Profiles Define Murine and Human MLL-AF4 Leukemias. Cancer Cell, 2008. 14: p. 355-368. 44. Okada, Y., et al., hDOT1L links histone methylation to leukemogenesis. Cell, 2005. 121(2): p. 167- 78. 45. Bernt, K.M., et al., MLL-rearranged leukemia is dependent on aberrant H3K79 methylation by DOT1L. Cancer Cell, 2011. 20(1): p. 66-78. 46. Deshpande, A.J., et al., AF10 regulates progressive H3K79 methylation and HOX gene expression in diverse AML subtypes. Cancer Cell, 2014. 26(6): p. 896-908. 47. Daigle, S.R., et al., Selective killing of mixed lineage leukemia cells by a potent small-molecule DOT1L inhibitor. Cancer Cell, 2011. 20(1): p. 53-65. 48. He, N., et al., Human Polymerase-Associated Factor complex (PAFc) connects the Super Elongation Complex (SEC) to RNA polymerase II on chromatin. Proc Natl Acad Sci U S A, 2011. 108(36): p. E636-45. 49. Biswas, D., et al., Function of leukemogenic mixed lineage leukemia 1 (MLL) fusion proteins through distinct partner protein complexes. Proc Natl Acad Sci U S A, 2011. 108(38): p. 15751-6. 50. Lu, H., et al., Gene target specificity of the Super Elongation Complex (SEC) family: how HIV-1 Tat employs selected SEC members to activate viral transcription. Nucleic Acids Res, 2015. 43(12): p. 5868-79. 51. So, C.W., et al., Dimerization contributes to oncogenic activation of MLL chimeras in acute leukemias. Cancer Cell, 2003. 4(2): p. 99-110. 52. Liedtke, M., et al., Self-association mediated by the Ras association 1 domain of AF6 activates the oncogenic potential of MLL-AF6. Blood, 2010. 116(1): p. 63-70. 53. Deshpande, A.J., et al., Leukemic transformation by the MLL-AF6 fusion oncogene requires the H3K79 methyltransferase Dot1l. Blood, 2013. 121(13): p. 2533-41. 54. Armstrong, S.a., et al., MLL translocations specify a distinct gene expression profile that distinguishes a unique leukemia. Nat Genet, 2002. 30: p. 41-47. 55. Faber, J., et al., HOXA9 is required for survival in human MLL-rearranged acute leukemias. Blood, 2009. 113(11): p. 2375-85. 56. Roychoudhury, J., et al., MEIS1 regulates an HLF-oxidative stress axis in MLL-fusion gene leukemia. Blood, 2015. 125(16): p. 2544-52. 57. Kroon, E., et al., Hoxa9 transforms primary bone marrow cells through specific collaboration with Meis1a but not Pbx1b. EMBO J, 1998. 17(13): p. 3714-25. 58. Zeisig, B.B., et al., Hoxa9 and Meis1 are key targets for MLL-ENL-mediated cellular immortalization. Mol Cell Biol, 2004. 24(2): p. 617-28. 59. Barabe, F., et al., Modeling the initiation and progression of human acute leukemia in mice. Science, 2007. 316(5824): p. 600-4. 60. Drynan, L.F., et al., Mll fusions generated by Cre-loxP-mediated de novo translocations can induce lineage reassignment in tumorigenesis. EMBO J, 2005. 24(17): p. 3136-46. 61. Horton, S.J., et al., MLL-AF9-mediated immortalization of human hematopoietic cells along different lineages changes during ontogeny. Leukemia, 2013. 27: p. 1116-26.

57

62. Chen, W., et al., Modeling human infant MLL leukemia in mice: leukemia from fetal liver differs from that originating in postnatal marrow. Blood, 2011. 117: p. 3474-3475. 63. Wei, J., et al., Microenvironment Determines Lineage Fate in a Human Model of MLL-AF9 Leukemia. Cancer Cell, 2008. 13: p. 483-495. 64. Liu, H., et al., Proteasome inhibitors evoke latent tumor suppression programs in Pro-B MLL Leukemias through MLL-AF4. Cancer Cell, 2014. 25: p. 530-542. 65. Jansen, M.W.J.C., et al., Immunobiological diversity in infant acute lymphoblastic leukemia is related to the occurrence and type of MLL gene rearrangement. Leukemia, 2007. 21: p. 633-641. 66. Aoki, Y., et al., Identification of CD34+ and CD34- leukemia-initiating cells in MLL-rearranged human acute lymphoblastic leukemia. Blood, 2015. 125(6): p. 967-80. 67. Stam, R.W., et al., Gene expression profiling-based dissection of MLL translocated and MLL germline acute lymphoblastic leukemia in infants. Blood, 2010. 115: p. 2835-2844. 68. Andersson, A.K., et al., The landscape of somatic mutations in infant MLL-rearranged acute lymphoblastic leukemias. Nat Genet, 2015. 47: p. 330-337. 69. Stumpel, D.J.P.M., et al., Specific promoter methylation identifies different subgroups of MLL- rearranged infant acute lymphoblastic leukemia, influences clinical outcome, and provides therapeutic options. Blood, 2011. 114: p. 5490-5498. 70. Kamezaki, K., L.L. Luchsinger, and H.-W. Snoeck, Differential requirement for wild-type Flt3 in leukemia initiation among mouse models of human leukemia. Exp Hematol, 2014. 42: p. 192- 203.e1. 71. Ayton, P.M. and M.L. Cleary, Transformation of myeloid progenitors by MLL oncoproteins is dependent on Hoxa7 and Hoxa9. Genes Dev, 2003. 17(18): p. 2298-307. 72. Kumar, A.R., et al., Hoxa9 influences the phenotype but not the incidence of Mll-AF9 fusion gene leukemia. Blood, 2004. 103(5): p. 1823-8. 73. So, C.W., et al., Leukemic transformation of hematopoietic progenitors by MLL-GAS7 in the absence of Hoxa7 or Hoxa9. Blood, 2004. 103(8): p. 3192-9. 74. Trentin, L., et al., Two independent gene signatures in pediatric t(4;11) acute lymphoblastic leukemia patients. Eur J Haematol, 2009. 83(5): p. 406-19. 75. Nishimoto, N., et al., Loss of AML1/Runx1 accelerates the development of MLL-ENL leukemia through down-regulation of p19ARF. Blood, 2011. 118(9): p. 2541-50. 76. Goyama, S., et al., Transcription factor RUNX1 promotes survival of acute myeloid leukemia cells. J Clin Invest, 2013. 123(9): p. 3876-88. 77. Montes, R., et al., Enforced expression of MLL-AF4 fusion in cord blood CD34+ cells enhances the hematopoietic repopulating cell function and clonogenic potential but is not sufficient to initiate leukemia. Blood, 2011. 117(18): p. 4746-58. 78. Ford, A.M., et al., In utero rearrangements in the trithorax-related oncogene in infant leukaemias. Nature, 1993. 363(6427): p. 358-60. 79. Menendez, P., et al., Bone marrow mesenchymal stem cells from infants with MLL-AF4+ acute leukemia harbor and express the MLL-AF4 fusion gene. J Exp Med, 2009. 206: p. 3131-3141. 80. Bueno, C., et al., A human ESC model for MLL-AF4 leukemic fusion gene reveals an impaired early hematopoietic-endothelial specification. Cell Res, 2012. 22(6): p. 986-1002. 81. Driessen, E.M., et al., Frequencies and prognostic impact of RAS mutations in MLL-rearranged acute lymphoblastic leukemia in infants. Haematologica, 2013. 98(6): p. 937-44. 82. Taketani, T., et al., FLT3 mutations in the activation loop of tyrosine kinase domain are frequently found in infant ALL with MLL rearrangements and pediatric ALL with hyperdiploidy. Blood, 2004. 103(3): p. 1085-8. 83. Tamai, H., et al., Activated K-Ras protein accelerates human MLL/AF4-induced leukemo- lymphomogenicity in a transgenic mouse model. Leukemia, 2011. 25(5): p. 888-91.

58

84. Prieto, C., et al., Activated KRAS Cooperates with MLL-AF4 to Promote Extramedullary Engraftment and Migration of Cord Blood CD34+ HSPC But Is Insufficient to Initiate Leukemia. Cancer Res, 2016. 76(8): p. 2478-89. 85. Bueno, C., et al., FLT3 activation cooperates with MLL-AF4 fusion protein to abrogate the hematopoietic specification of human ESCs. Blood, 2013. 121(19): p. 3867-78, S1-3. 86. Montes, R., et al., Ligand-independent FLT3 activation does not cooperate with MLL-AF4 to immortalize/transform cord blood CD34+ cells. Leukemia, 2014. 28(3): p. 666-74. 87. Bursen, A., et al., The AF4.MLL fusion protein is capable of inducing ALL in mice without requirement of MLL.AF4. Blood, 2010. 115(17): p. 3570-9. 88. Kowarz, E., et al., Complex MLL rearrangements in t(4;11) leukemia patients with absent AF4.MLL fusion allele. Leukemia, 2007. 21(6): p. 1232-8. 89. von Bergh, A., et al., Cryptic t(4;11) encoding MLL-AF4 due to insertion of 5' MLL sequences in . Leukemia, 2001. 15(4): p. 595-600. 90. Kumar, A.R., et al., t(4;11) leukemias display addiction to MLL-AF4 but not to AF4-MLL. Leuk Res, 2011. 35(3): p. 305-9. 91. Kalberer, C.P., J. Antonchuk, and R. Keith Humphries, Genetic modification of murine hematopoietic stem cells by . Methods Mol Med, 2002. 63: p. 231-42. 92. Skalamera, D., et al., Generation of a genome scale lentiviral vector library for EF1alpha promoter-driven expression of human ORFs and identification of human genes affecting viral titer. PLoS One, 2012. 7(12): p. e51733. 93. Forster, A., et al., Engineering de novo reciprocal chromosomal translocations associated with Mll to replicate primary events of human cancer. Cancer Cell, 2003. 3(5): p. 449-58. 94. Metzler, M., et al., A conditional model of MLL-AF4 B-cell tumourigenesis using invertor technology. Oncogene, 2006. 25(22): p. 3093-103. 95. Li, S., et al., The P190, P210, and P230 forms of the BCR/ABL oncogene induce a similar chronic myeloid leukemia-like syndrome in mice but have different lymphoid leukemogenic activity. The Journal of experimental medicine, 1999. 189: p. 1399-1412. 96. Chen, W., et al., A murine Mll-AF4 knock-in model results in lymphoid and myeloid deregulation and hematologic malignancy. Blood, 2006. 108(2): p. 669-77. 97. Burmeister, T., et al., The MLL recombinome of adult CD10-negative B-cell precursor acute lymphoblastic leukemia: Results from the GMALL study group. Blood, 2009. 113: p. 4011-4015. 98. Hystad, M.E., et al., Characterization of early stages of human B cell development by gene expression profiling. J Immunol, 2007. 179(6): p. 3662-71. 99. Geng, H., et al., Self-enforcing feedback activation between BCL6 and pre-B cell receptor signaling defines a distinct subtype of acute lymphoblastic leukemia. Cancer Cell, 2015. 27(3): p. 409-25. 100. Wilkinson, A.C., et al., RUNX1 is a key target in t(4;11) leukemias that contributes to gene activation through an AF4-MLL complex interaction. Cell Rep, 2013. 3(1): p. 116-27. 101. Wang, Q.F., et al., MLL fusion proteins preferentially regulate a subset of wild-type MLL target genes in the leukemic genome. Blood, 2011. 117(25): p. 6895-905. 102. Goyama, S., M. Wunderlich, and J.C. Mulloy, Xenograft models for normal and malignant stem cells. Blood, 2015. 125: p. 2630-2641. 103. Topp, M.S., et al., Targeted therapy with the T-cell-engaging antibody blinatumomab of chemotherapy-refractory minimal residual disease in B-lineage acute lymphoblastic leukemia patients results in high response rate and prolonged leukemia-free survival. J Clin Oncol, 2011. 29(18): p. 2493-8. 104. Rayes, A., R.L. McMasters, and M.M. O'Brien, Lineage Switch in MLL-Rearranged Infant Leukemia Following CD19-Directed Therapy. Pediatr Blood Cancer, 2016. 63(6): p. 1113-5.

59

105. Gardner, R., et al., Acquisition of a CD19 negative myeloid phenotype allows immune escape of MLL-rearranged B-ALL from CD19 CAR-T cell therapy. Blood, 2016. 106. Chen, W., et al., Malignant Transformation Initiated by Mll-AF9: Gene Dosage and Critical Target Cells. Cancer Cell, 2008. 13: p. 432-440. 107. Kamps, M.P. and D. Baltimore, E2A-Pbx1, the t(1;19) translocation protein of human pre-B-cell acute lymphocytic leukemia, causes acute myeloid leukemia in mice. Mol Cell Biol, 1993. 13(1): p. 351-7. 108. Luo, R.T., et al., The elongation domain of ELL is dispensable but its ELL-associated factor 1 interaction domain is essential for MLL-ELL-induced leukemogenesis. Mol Cell Biol, 2001. 21(16): p. 5678-87. 109. Ptasinska, A., et al., Identification of a dynamic core transcriptional network in t(8;21) AML that regulates differentiation block and self-renewal. Cell Rep, 2014. 8(6): p. 1974-88. 110. Huber, W., et al., Orchestrating high-throughput genomic analysis with Bioconductor. Nat Methods, 2015. 12: p. 115-121. 111. Robinson, M.D., D.J. McCarthy, and G.K. Smyth, edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics, 2010. 26: p. 139-40. 112. Subramanian, A., et al., Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A, 2005. 102(43): p. 15545- 50. 113. Sartor, M.A., G.D. Leikauf, and M. Medvedovic, LRpath: A logistic regression approach for identifying enriched biological groups in gene expression data. Bioinformatics, 2009. 25: p. 211- 217.

60

2.6 Appendix Tables

Table 1. PCR primers

Realtime QPCR primers CSF1R CAGAGGAGAGGAACGTGTGT IKZF1 CATGGATGCTGATGAGGGTCA GGACACTGGGCTCTATCACT GCCCTCATCTGGAGTATCGC CSF3R TGGAGGATGGAACAGAATGGG IRF4 TTGAGGAACTGGTTGAGCGGA CAGGTCTTGCCAATGTGCTTT AGGGGTGGCATCATGTAGTTGT GATA2 CCAGCTTCACCCCTAAGCAG TCF3 GAGAATGAACCAGCCGCAGA TTCGCTTGGGCTTGATGAGT CCTTCCCGTTGGTGACAGG GAPDH AGCCACATCGCTCAGACAC TTAAAAGCAGCCCTGGTGAC RT-PCR primers MLL-Af4 GATGGAGTCCACAGGATCAGA PGK GGGAAAAGATGCTTCTGGGAA GCTGGAGCTGCTCTCACTCTCA TTGGAAAGTGAAGCTCGGAAA ChIP QPCR primers HOXA9 ATGCTTGTGGTTCTCCTCCAGTT IGF2BP3 GACCACGAACGGGAGAACTG CCGCCGCTCTCATTCTCAGC TCAATTCAGACGTGGTGCGG G HOXA10 CGCAACCACCCCAGCCAG CCNJ GCGCTTTGGCAACTCAGG TTGTCCGCCGAGTCGTAGAGG CAGGAAGAGCCGAAGGACTC MEIS1 TTTGGTAGTTGGGTCTGAGGGG BMI1 GAACCAAGGAGAAAGCGCC GATCTTCTCTCCCCCAAAAATCA TCCAGCTCTCGCTTGTCCAG RUNX1 TGCAGAAGTTCCGTCGCT ZNF521 GAGCCTGGTCCACCGTTAC G GCAACAGCCAGAAACGGC TCTGGGGGAAACCTAGACCC IVL GCCGTGCTTTGGAGTTCTTA KLRF2 AGATCATGCCACTGCACTC CCTCTGCTGCTGCCACTT GTGGGCTTTAGGGTCTGAATAA PROM1 TTTCGGGGAAACTGATCCCG COL19A1 GGAAGTTTGTTGGCCCGGAA CGACCGGACAACAAAGAGGA CCTCTAAGGGGGTTCGCTTAC GNAQ GGTGCAGACGAGAATAACGC IKZF2 CTCCCACCTACTGGTCTCCG AGGGCAAGCCCAAATGTGTC CTTATCCCGCTTCAGGTCCC TACTGCGACGAGGAGGAGAA GGCAGCAGCTCGAATTTCTT

61

Table 2. MLL-AF4 ALL signature from patient dataset

Stam MLL-AF4 vs Stam MLL-AF4 vs Andersson MLL-AF4 vs Andersson MLL-AF4 vs

non-MLLr Up non-MLLr Down non-MLLr Up non-MLLr Down

TSPAN AAED1 KCNE1L S100A10 AKAP12 KIT ADAM10 LAMP5 AFAP1L2 DIP2C HLA-B MTX3 RGS3 9

ABHD17C KCNK12 SACS ALDH1A1 KLF11 AGO4 LATS2 AGPAT3 DLEU7 HRK MYO1B RHOF TSPYL5

TUBA4 ABHD4 KLRC4 SAMHD1 ANGPT2 KLF8 ALDH9A1 LGALS8 AHNAK2 DPEP1 ICAM3 MYO5C RLTPR A

LOC10012993 ABR KLRK1 SDC2 ANKRD33B KLRB1 ANGEL2 ALDH2 DTNB ID3 MZB1 RUNX3 TUSC1 1

UBE2E ALOX5AP LAMP5 SETD6 ARHGAP21 LAMA5 AP3M1 LOC339862 ALOX5 DTX1 IGLL1 MZT2A RYK 3

ANKRD30 ANXA2 LATS2 SGK1 BEX4 LCK ARNTL MAN2C1 DUSP16 IGSF3 N4BP3 S1PR4 USP28 B

USP6N ASAP2 LCN8 SIT1 C10orf10 LIG4 ATG13 MEAF6 ANO9 IKZF2 NDFIP1 SAMD4A L

LOC1001 ATP8B4 LGALS1 SKIDA1 CA2 ATG4C MEIS1 APP IQSEC1 NEIL1 SARDH VIPR2 29447

LOC1002 BAALC LHFP SLC35G2 CD24 ATP6V0A2 METTL7A AQP5 EGLN1 ITIH3 NEURL SCML1 VOPP1 88685

LOC1002 ARHGAP2 BANK1 LILRB1 SMC6 CD27 ATR MLH3 EHD1 ITPR3 NFATC4 SCRN1 WBP5 90036 1

LOC1005 ARHGAP2 BMP3 LIPC SMIM24 CD3D BDH1 MPPE1 ELFN2 JAG2 NINJ1 SEMA3F YES1 05801 4

LOC1005 ARHGAP3 BRE-AS1 LMO2 SNHG4 CD52 BRE MPZL1 ELK3 JUP NKAIN4 SEMA4B ZEB1 07185 2

LOC10012 LOC6524 ZEB1- CACNA2D4 SOCS2 CERS4 BRPF1 NFATC3 ARHGEF5 ELOVL2 KCNA6 NLRC3 SEMA6C 8868 93 AS1

LOC10013 SOCS2- CCNA1 CMTM7 ADGRL1 C11orf24 NOL8 ATN1 ENPP4 KCND1 NOTCH2 SERPINB9 ZNF135 0458 AS1

LOC10028 CD72 SPRY2 CMTM8 LSP1 C5orf24 OSGEPL1 BCL6B EPB41L5 KCNJ16 NOTCH3 SH2D3A ZNF256 8781

LOC10050 ZNF354 CD93 STAR CNFN MAGED1 C8orf82 PAN3 BCORL1 EPHB6 KCNJ2 NRG3 SH2D4B 5956 C

LOC33986 KCNJ2- CDC42EP3 STX1A CNN3 MARCKS CCDC132 PCDHGC3 BEST3 EXD2 NRP1 SH3TC1 ZNF467 2 AS1

LOC44086 CEBPA TMEM71 COL5A1 MDK CCNH PCDHGC5 BEX4 FAM102A KCNQ1 NSUN7 SIAH2 ZNF608 4

62

LOC64307 CECR6 TPPP3 CORO2B MME CDK6 PIGF BTG2 FAM129A KCTD17 NUDT11 SIDT1 ZNF662 2

LOC72806 KHDRBS NUP62C CENPV UCK2 CSF2RB MS4A1 CLN3 POGLUT1 C10orf118 FAM213A SLC16A14 ZNF827 1 3 L

CLEC14A LPAR4 VAT1L CXCL3 MYBL2 COG7 PREB C12orf49 FAM213B KHNYN NYNRIN SLC43A2 ZNF90

FAM41AY CMTM4 LRRC32 VCAN CYFIP2 MZB1 COQ2 PTER C14orf132 KIT OCIAD2 SLC48A1 2

CORO1C MATR3 VLDLR CYTL1 NDFIP1 DCAF8 RB1 C1R FAM65A KLHL33 OR2A20P SLC5A3

NGFRAP CPEB2 MCEMP1 VNN1 DDR1 DDX59 RER1 C1RL-AS1 FAM65B LAMA5 ORMDL3 SLC6A6 1

LAPTM4 CRISPLD1 MEIS1 WASIR1 DEF8 NID2 DFFB RNF220 C2CD2 FAM84B PAG1 SLFN13 B

LDLRAD CSPG4 METTL7A WASIR2 DPEP1 NKAIN4 DIMT1 RSAD1 C3orf67 FBLN1 PALD1 SLFNL1 3

CSRP2 MLC1 WIPI1 DSG2 NRG3 DNAJC24 SCPEP1 C4orf32 FBXO18 LIG4 PARD3 SMAD1

LINC004 CXXC5 MN1 WT1 DTX1 NRN1 DRAM2 SCRN3 CAMK1D FCGR3B PAWR SMAD7 94

LOC1004 DAD1 MOK ZC3H12C ELK3 OCIAD2 DTWD1 SETD6 CARNS1 FCRL2 PCGF5 SMO 22737

LOC1151 DEPDC7 NADK2 ZCCHC7 EPSTI1 P2RY14 DUSP3 SGK223 CASP7 FCRL5 PDK4 SOD2 10

PHACTR LOC2830 DHRS3 NFIL3 FAM134B EIF4G3 SLC27A1 CBLN3 FHIT PDXK ST3GAL5 1 70

PLEKHG LOXL1- DYNC2LI1 NLRP3 FAM19A5 EMB SNAI3 CCDC85C FLJ16779 PHGDH STARD9 4B AS1

ECRP NOG FAM213A PRKCH EML2 SNX2 CD24 FZD6 LRBA PITPNM2 STIM2

EMB NR4A2 FAM213B PTP4A3 ERP44 SRD5A3 CD2AP GALM LRIG1 PLCL2 STK33

EMBP1 NR5A2 FAT1 REM2 FAM175A TBC1D14 CD4 GALNT11 LRP5L PLD4 SYNE2

PLEKHG EMP1 NRXN2 FCRL1 RUNX3 FAM204A TMEM218 CD52 GALNT4 LRP6 SYTL1 4B

LRRC14 FAM69C P2RX5 FHIT S1PR1 FAM206A TMX2 CD9 GCHFR PON2 TBC1D10C B

FAM78A PAN3 GCHFR SEMA6A FAM78A TPM4 CDC42EP4 GDPD5 LSP1 PPARD TBL1X

SERPINB FCGR2B PAN3-AS1 GIMAP7 FBXW2 TRMT13 CDKN1C GFOD1 LTB PPFIBP1 TCF4 9

FCRLB PCDHGA1 GNAI1 SETBP1 FLT3 TRPM7 CELSR1 GGT1 MAGED1 PRRT2 TCF7L2

FLT3 PDGFRB GPER1 SHANK3 GALC TSPAN31 CERK GNA11 MAML2 PSEN2 TDRKH

SLC16A1 FOSL2 PIWIL4 GSAP GANC UBASH3B CLEC2B GNG7 MAN1A1 PTBP3 TET3 4

63

FZD1 PPP1R14A GZMB SMAD1 GCNT1 UCK2 CLSTN3 GPHN MAP2K3 PTP4A3 TIFA

GAS2L3 PROM1 HCP5 STK32B GFM2 UFM1 CMTM7 GPR110 MARCKS PTPRD TJP2

GRAMD1 GPM6B PTGER2 HOXB2 TCF4 GNAQ USP48 CMTM8 MESDC1 PXN TLE4 B

GPR183 PTH1R HPGD TCL1B GOPC VPS36 COBL GRIK5 MEX3A RAB32 TM7SF2

GREM1 PTPN14 ICAM3 TIFA GOSR2 VPS39 COL18A1 GSDMB MFAP2 RAB3IP TMEM63C

TMEM17 TNFRSF13 HTRA3 RAB17 IFITM1 HENMT1 VPS52 COL5A1 GSPT2 MFHAS1 RAB43 6A C

MGC393 IGFBP7 RBKS IGHM TRAT1 HEXB WDSUB1 CYFIP1 GYLTL1B RAG1 TNFRSF14 72

IL10RB-AS1 RBM24 JCHAIN TRBC1 HPS3 XYLT1 DAPK1 HAP1 MME RAI14 TNFRSF21

IRGM RGS16 IGK TRBC2 HUS1 YIPF4 DBH HDAC7 MNF1 RANBP6 TNK2

IGKV1OR2- RAPGEF IRX3 RNASE2 YES1 IDH1 ZEB2 DDR1 HECTD2 MREG TOP3B 108 3

ITPRIPL2 RNASE3 IGLL1 ZEB1 IGF2BP2 ZNF354B DEF8 HIP1R MS4A1 RASL12 TRIM47

JMJD1C-AS1 RNASE6 KHDRBS3 ZNF329 IKZF1 DGKZ HIST1H1C MSR1 RASSF8 TRIM6

MTERFD KCNA5 RRAS2 KIAA0226L ZNF827 KRI1 DHCR7 HIST4H4 RGPD5 TSHR 3

64

Table 3. proB and preB signature Hystad ProB signature (ProB vs PreB Up) Hystad PreB signature(PreB vs ProB Up) ABLIM1 HMGB1 PPM1F ADAM19 IL13RA1 ADA HOXA9 PRKCH APBB2 IL4R AKAP12 HPS4 PKN2 ARHGEF3 IRF4 APOL3 HSPA1A PRNP RHOH IRS1 ARHGEF10 HSP90AA1 PROM1 ARR3 SPIDR ARID5B HTATSF1 PSIP1 B4GALT1 DTX4 BCL11A IARS2 PTGER2 BASP1 KLHL14 BCLAF1 IDUA RAP1GAP2 FAM129C LAMA5 BCR IFIT2 RBM26 BET1L LYN BLNK IL1B RCSD1 BIK SMAD7 BST1 IQGAP2 RNASET2 BIRC3 MAPRE1 BZRAP1-AS1 ITGA6 RRAGD BTG1 MARCKS CCND2 KDSR SERPINB1 KIAA0226L NEK2 CD34 LCP2 SIAH2 LAMP5 NFATC4 CD99 LHFP SLC2A5 CACNA1A NT5E CDH2 LIG4 SLC38A2 CAMK2D P2RX5 CDK6 LILRB1 SOCS2 CAMK4 P2RY10 CDKN1B LMO2 SOD1 CD19 PTPN6 CEP68 LYPLAL1 SORL1 CD72 PTPRC CHD1 MTF2 STAT5A CD86 RGS16 CKLF MAP4K3 SUSD3 CDC25B RGS2 COMMD3 MCFD2 TCF7L2 CXCR4 RNGTT CRIP2 MEIS1 KLF10 CYBB RPS6KB2 CTGF MOB3B TM7SF3 DCXR SH3BP5 YOD1 MSH6 TNFSF4 DTX1 SIK1 DNTT MYC TOP2B SIPR1 FCRL2 EBF1 NEAT1 TP53INP1 ETS1 SPIB EFNA1 NEDD4 TPT1 FCRL1 TAGAP ERG NOTCH1 TSC22D1 FCRL3 TCL1A ETS2 NR3C1 XBP1 ATHL1 TFDP2 EVI2A NRP1 ZFP36L2 FCRLA TNFRSF21 FHIT OAS3 ZMYM2 GPR18 TOM1L1 FYN LPAR6 HCK TRIB2 GFI1 PAG1 HDAC9 ANKRD36B GLRX PAICS SYVN1 FAM69A GNAQ PAM HRK FLJ45513 GNG10 PCM1 IFNGR2 PRDM11 GNG11 PDE4B IGF2R ZMAT3 GSN PECAM1 IGFBP3 VDR HHEX PLA2G6 IGHM VNN2 MDFIC PLK2 JCHAIN ZFP36 HIVEP2 TMEM123 IGLJ3 ZKSCAN7

65

Table 4. Andersson 100-gene discriminator

AAK1 DTX3 LOC100129931 RUNX1 ACADSB EDA2R LRRC61 RUNX3 ADCY6 EEF2K LY9 SCML1 ADD2 FAM66B MEF2A SCML2 ALS2 FCHO2 METRN SENP6 AMPD2 FGFR1 MMRN1 SERPINB9 BMI1 GALM MOCS2 SLFN12L C14orf37 GALNT2 MYO6 SNCB CALML4 GCAT NADK2 SPAG16 CASC10 GIPR NFATC4 SPAG6 CARNS1 GPRASP1 NLRC3 SRD5A3 CBLN3 GYLTL1B NRXN3 SYNGR1 CCDC85C HK2 OR2A9P TBC1D14 CD248 HSD17B12 P2RX5 TBC1D16 CDC42EP4 IARS PALM TNFRSF13B CDK8 IGF2BP2 PAN3 TOX2 CEP112 IGIP PECR TPD52L2 CLEC14A KCND1 PENK TTC12 CLN6 KCNQ1 PIP4K2C UBASH3B COLCA1 KHDRBS3 PLXDC2 USP45 CRMP1 LATS2 PPM1H ZBED6CL DACH1 LCN6 PPP1R3G ZMAT3 DHRS3 LDLRAD3 PTPRK ZMYND10 DNAJC5 LHFP PURA ZNF423 DPF3 LOC100128593 RPP40 ZNF521

66

Table 5. Differential signature genes between homemade ALL of MLL-Af4 and MLL-AF9 AADAT C1orf43 CTNND1 GALNT3 KIAA1328 MGAT5 OR2W3 RAMP1 SOCS2 TSC22D1 AAK1 C20orf27 CTSZ GCSAM KIF13A MIR501 OXSM RAPGEF SOD2 TSEN2

ABCD3 C22orf29 CXXC5 GGT7 KIF26B 0MLC1 P4HA2 5RCN3 SPAG1 TSHZ1 ABCG1 C7orf50 CYB561 GIPC3 KLF12 MLH3 PABPC4 REEP3 SPAG16 TSNARE1 ABLIM1 C7orf57 CYB561D GIPR KLF2 MLKL PAPLN RELB SPAG6 TSPO

ADAT2 CA5BP1 2CYB5R4 GMPPB KLHL12 MLLT3 PCDH9 REPS2 SPATA20 TSTD1 ADCY9 CAP2 CYBRD1 GPM6B KLLN MME PCED1B RFX5 SPIB TTC12 ADD2 CARD6 CYGB GPR173 KLRB1 MMP17 PCOLCE RHOB SPINT1 TTC21B AEN CASK DDN GRAMD4 L3HYPDH MN1 PDE3B RIC8B SPP1 TTC7A AGTRAP CASZ1 DFNB31 GREM1 LBH MPEG1 PDHA1 RNF144 SRGN TTPAL

AKAP1 CBX2 DLEU7- GSPT1 LHPP MPI PENK ARNFT2 SRM UBE3C AMICA1 CCDC17 AS1DOCK9 GYG1 LINC00467 MPP7 PEX5 ROR1 SRPRB UBL3 ANGEL1 1CCDC64 DOK4 HCK LINC00667 MPZL3 PFAS RRNAD1 SRSF8 UGCG ANKRD6 CCDC78 DSE HHIP-AS1 LIPE MRI1 PHLDB2 RTKN2 SSH2 USF1 AOX2P CCDC85 DTX4 HIST1H1E LMO7 MTA3 PHLPP1 RUFY3 ST6GAL1 VAT1L

APCDD1 CCCDC92 DUSP16 HIST1H2B LOC1004227 MTL5 PIM1 RUNX1 STIM2 VWCE ARF3 CCNG2 DUSP3 HHIST1H4J 37LOC1004994 MTOR PIP4K2C RWDD2 STK32B WBSCR27 ARFIP1 CCNJ EFNB1 HMBS 89LOC1005061 MXRA PLEK ARXRA STRBP WDFY4 ARHGAP CD160 EGR3 HPCAL1 24LOC1005065 7MYC PLOD1 SBK1 STT3B WDR43 12ARHGEF1 CD180 EIF2AK4 ICA1 85LOC151174 MYEF2 PLTP SCARF1 SYNGR1 WDR62 1ARHGEF6 CD24 EIF3A ICAM2 LOC283710 MYL4 PLXND1 SCLT1 SYNPO WWP1 ARL4A CD244 ELL3 ID2 LOC648987 MYO1 PNMA1 SCML2 SYTL3 XXYLT1

ARL4C CD34 ELOVL2 IGSF3 LOC729732 GNCKAP PPARGC1 SEMA4A TBC1D16 YIPF6 ARMC10 CDH24 EMC1 IKZF2 LOXL4 5NCLN BPPM1F SEMA4B TBKBP1 YWHAZ ARRB1 CDK2AP ENDOV IL4R LPAR5 NEGR1 PPP1R14C SEPW1 TBL1X ZC3H14

ATHL1 2CEP112 EXT2 IPO4 LRP11 NEURL PRDX2 SESN1 TCL1B ZC3HAV1 ATP13A2 CHST11 FAM110 IRF4 LRRC1 2NFATC PREX1 SFMBT2 TES LZDHHC23 ATP1A1 CHST7 AFAM129C IRF5 LRRC61 4NFIC PRICKLE SH2D2A TGFB1I1 ZDHHC9 ATP2A2 CLEC14 FAM168 IRS1 LRRC8C NID2 4PROM1 SH3BP2 THNSL2 ZFP28 ATP6V0A ACLEC17 AFAM171 ITGB7 LSP1 NMT2 PSD2 SKIDA1 TLR9 ZHX1 1ATP8B2 ACLEC2D A1FAM184 ITPRIPL2 LSP1P3 NOXA1 PSTPIP2 SLC20A2 TMEM121 ZHX2 B3GNT5 CLSTN1 AFAM46A ITSN1 LTBP2 NPTX2 PTGDR2 SLC25A4 TMEM185 ZHX3 B3GNT7 CLUH FBXO25 JAG2 LTBR NR3C1 PTGER4 3SLC2A6 BTMEM216 ZKSCAN7 BAALC CMAHP FCHO1 JAKMIP2 LYN NTN3 PTPN6 SLC30A4 TNFAIP8 ZNF146 BCL6 CNKSR2 FHOD3 JAZF1- LYRM5 NUAK2 PTPRB SLC30A5 TNFRSF11 ZNF256

BEX4 CNN2 FKBP5 AS1JDP2 MAPKAPK3 NUDT2 PTPRR SLC38A5 ATNFRSF1 ZNF320 BLNK COL1A1 FLVCR1- KCNA6 MAT2A 2OAF PTPRS SLC7A5 BTNNI2 ZNF423 BNC2 COL9A2 AS1FMO4 KCNQ1 MBLAC2 OFD1 PURA SLC9B2 TPD52L2 ZNF521 BTBD6 COMMD FOXK1 KHDRBS2 MBTPS2 OMG PWP2 SLFN13 TPK1 ZNF582

C10orf54 5CPEB3 FOXO3 KHDRBS3 MDGA1 OPALI PYGM SMAD2 TPST1 ZNF608 C16orf54 CRB2 FRMD4B KIAA0226 MED8 NOPRL1 QPRT SMYD5 TRIB3 ZNF703 C16orf74 CRIP1 FTSJ1 LKIAA0355 MFI2 OR2A1 RAB10 SNX10 TRIM65 ZNF836 C1orf216 CSMD1 FZD1 KIAA0922 MFNG OR2A7 RAB11FIP SNX30 TRPM2 ZUFSP

1

67

Chapter 3 FOXO1-induced self-renewal defines the AML1-ETO pre-leukemic program

3.1 Introduction

AML1-ETO acute myeloid leukemia

The t(8;21)(q22;q22) translocation is one of the most frequent genetic alterations in AML, accounting for

~12% of the disease[1]. Leukemic blasts with the t(8;21) translocation show granulocytic maturation and a distinct immunophenotype characterized by a prevalent positivity for CD19, CD56, CD13, and

CD34[2]. Molecular cloning of the translocation breakpoints revealed rearrangement of the AML1 gene

(also referred to as RUNX1) on chromosome 21q22 and the ETO gene (also referred to as RUNX1T1 or

MTG8) on chromosome 8q22[3-6], resulting in the AML1-ETO (AE) fusion protein. AML1 is a transcription factor critical for definitive hematopoiesis. AML1 knockout mice lacked fetal liver hematopoiesis and died at E11.5-E12.5 [7-10]. Progenitor activity was severely impaired in cells from the yolk sac and fetal liver. In contrast, conditional AML1 knockout mice showed that AML1 is dispensable for adult hematopoiesis since mice survived AML1 deletion[11, 12]. AML1 loss minimally impacts long- term HSC activity, and instead is associated with an expansion of myeloid progenitors, which is possibly due to the block of myeloid terminal differentiation[13]. ETO is a transcriptional repressor that is not normally expressed in hematopoietic cells. It interacts with several corepressors including nuclear corepressor (N-CoR), SMRT, mSin3A and the histone deacetylases (HDACs)[14]. Since the chromosomal breakpoints of the t(8;21) cluster within AML1 5 and ETO intron 1, this preserves the RUNT DNA binding domain of AML1, and replaces the transcription activation domain of AML1 by almost the entire ETO protein[5], suggesting that AE has dominant negative function repressing AML1 target genes. This view is supported by the demonstration that mice with heterozygous germline knock-in of AE phenocopied the AML1 knockout mice[15-17], of which embryonic lethality occurred around

E12.5-E13.5 with lack of definitive hematopoiesis. However, fetal livers of AE knock-in mice displayed dysplastic multilineage hematopoietic progenitors with an increased self-renewal capacity in vitro, not observed in AML1-/- mice[15], suggesting that AE has additional functions other than AML1 disruption. 68

In accordance, several studies suggest that AE protein has altered DNA binding properties compared to wildtype AML1. AE has a selective preference for certain target genes that contain multimerized AML1 consensus sites in their regulatory elements[18]. Additionally, AE binds to the canonical short AML1 motif (5’-TG(T/C)GGT-3’) more efficiently than AML1, whereas AML1 prefers a longer motif (5’-

TGTGGTTT-3’; with 2 additional thymidines to the short motif at the 3’ position) than AE[19]. This differential DNA binding preference may contribute to largely correlated but non-identical genome-wide distributions of AE and native AML1[20, 21], and thus different regulated target genes.

AE pre-leukemia cells

Leukemogenesis is a hierarchical process, with an initiating mutation establishing pre-leukemia stem cells that evolve over time to overt disease through additional cooperating mutations[22, 23]. Several studies suggest that pre-leukemia stem cells can survive chemotherapy and serve as a potential reservoir of disease relapse [24, 25]. Although t(8;21) AML has a comparatively good prognosis and most patients enter remission, half of patients relapse and only a 60% overall survival is achieved after five years [26].

Pre-leukemia stem cells are evident in this AML subtype, as cells positive for AE can be detected long before disease onset or after complete remission[27-29]. As introduced in Chapter 1, AE mouse models have demonstrated the necessity of a cooperating event for AE leukemogenesis. Complementing the mouse studies, our group and others have shown that expression of AE in human CD34+ HSPC causes dysregulated differentiation and increased self-renewal of cells but without inducing AML [30, 31], serving as an ideal model to study the pre-leukemia stage of t(8;21) AML (AE cells).

The mechanism whereby AE mediates pre-leukemia programming is being gradually revealed. AE can block differentiation by interfering with the function of other transcription factors via physical interactions. CEBPA and PU.1 are transcription factors that play crucial roles in myeloid differentiation.

AE interacts with them and reduces their DNA binding activity[32, 33]. GATA1 is a major erythroid transcription factor, and AE hampers transcriptional activity of GATA1 by preventing its acetylation[34].

In particular, several independent studies have shown that CEBPA is an important downstream target of

69

AE. Blasts in t(8;21) AML showed relatively low CEBPA expression compared with other subgroups of

AML. Experimentally, expression of AE in U937 cells induced downregulation of CEBPA mRNA through inhibition of positive autoregulation in the CEBPA promoter[32]. Conversely, depletion of AE resulted in CEBPA upregulation and establishes a differentiation-associated transcriptional network dominated by de novo binding of CEBPA[35]. Thus, CEBPA is a critical target gene repressed by AE, whose downregulation contributes to the AE-mediated block of myeloid differentiation.

How AE pre-leukemia cells acquire aberrant self-renewal is less defined. The WNT/β-Catenin pathway is required for self-renewal of MLL-AF9-mediated AML in mouse models[36]. Several AE downstream targets appear to have convergent effects to activate this pathway. It was reported that AE activated 2 (COX2), which in turn increase β-Catenin protein[37]. JUP (gamma-Catenin) is upregulated by AE, which can activate WNT targets similarly to β-Catenin, via interacting with β-Catenin coactivator TCF[38]. Another enhancer of WNT signaling, groucho-related N-terminal enhancer of split

(AES), is reported to be upregulated in AE positive AML[39]. Additionally, AE represses secreted- related protein 1(SFRP1), an antagonist of WNT[40]. Accordingly, deletion of β-Catenin or overexpression of a dominant negative form of TCF, knockdown of AES and overexpression of SFRP1 impair proliferation and clonogenic capacity of AE expressing cells[37, 39, 40], indicating the WNT/beta-

Catenin pathway is also critical for self-renewal maintenance of AE leukemia[37, 41]. In addition to

WNT signaling, we and others have reported thrombopoietin (TPO) signaling is critical for maintaining survival and self-renewal of AE-expressing cells. MPL is the receptor of TPO and highly expressed in t(8;21) AML. Enhanced TPO/MPL signaling led to upregulation of anti-apoptotic protein Bcl-xL and activated PI3K/AKT and JAK/STAT pathways[42, 43], although how this results in increased self- renewal is unclear. Continued efforts to understanding the molecular mechanism mediating aberrant self- renewal of AE pre-leukemia cells will provide therapeutic opportunities for eliminating these stem cells and preventing disease relapse.

70

FOXO transcription factors.

The FOXO transcription factors comprise four members, FOXO1, FOXO3, FOXO4 and FOXO6.

FOXO1, 3 and 4 are ubiquitously expressed in various tissues, while FOXO6 is less conserved with the other three and expression is restricted to the central nervous system[44, 45]. Most studies focus on the first three FOXO members. Though FOXO1,3 and 4 have conserved protein domains and similar structure, mouse genetics suggest they have non-redundant functions. Foxo1-/- mice are embryonic lethal likely due to vascular development defects. Foxo3-/- mice can survive into adulthood, however, female mice are infertile with abnormal ovarian follicle development. In contrast, Foxo4-/- mice do not present any obvious abnormalities[46]. In addition to transcription regulation, the activity of FOXO factors are finely controlled by post-translational modifications. The best known inhibitory regulation is AKT- mediated phosphorylation[47]. Once phosphorylated by AKT, FOXOs can be captured by 14-3-3 proteins and exported from the nucleus to the cytoplasm[48]. Cytoplasmic FOXOs can reenter the nucleus upon proper reversed stimuli, or are ubiquitinated and further degraded by proteasome[49]. FOXOs can also be negatively regulated by a number of including SGK, IKK β and ERK[50-52]. By contrast, phosphorylation-mediated activation of FOXOs exists. In response to oxidative stress conditions, JNK phosphorylates FOXO4 and induce its translocation from the cytoplasm to the nucleus[53]. Besides phosphorylation, other posttranslational modifications of FOXOs, such as methylation and acetylation, have been reported, revealing the complexity of FOXO regulation[54].

FOXOs and tumorigenesis

FOXOs are generally considered as tumor suppressors. Somatic deletion of all Foxo1,3 and 4 in mice resulted in lymphoma and hemangioma development. The tumor formation has not been observed in any single Foxo-deficient mice, and was attenuated or absent in mice with compound mutations but retaining one intact Foxo allele, indicating the tumor suppressor function of FOXOs are redundant [55]. Possibly due to this redundancy, loss of function mutations of FOXO genes are not frequently present in human tumors, although the deletion involving chromosome 13q14, where FOXO1 resides, has been found in

71

~30% of prostate cancer samples. Ectopic expression of FOXO1 in prostate cancer cells inhibited their proliferation and survival[56]. Compared to genetic ablation, post-translational modifications could be a more efficient way to repress all FOXOs activities in cancer cells. AKT, ERK and IKK signaling pathways are often hyper-activated in by oncogenes, gain-of-function mutations in pathway components or loss of negative regulator of the signaling. Numerous studies have reported that FOXOs activities are suppressed in various types of cancer cells, and restoration of FOXOs activities is detrimental to the growth of these cancers. For instance, overexpression of NPM-ALK, an oncofusion protein resulting from t(2;5)(p23;q35) and prevalent in anaplastic large-cell lymphoma, led to concomitant AKT activation and phosphorylation of FOXO3 in Ba/F3 cells. Introducing a constitutively active form of FOXO3 that cannot be phosphorylated by AKT inhibited the proliferation of NPM-ALK- transformed Ba/F3 cells[57]. Genetic ablation of both AKT1 and AKT2 resulted in markedly reduced proliferation and metastasis of human colon cancer cell lines, associated with the absence of phosphorylated FOXOs[58]. In a breast cancer cell line, it was shown that ERK promoted tumorigenesis partially through phosphorylation and degradation of FOXO3[52]. FOXO3 activity is repressed by both

AKT and ERK signaling in patient derived glioblastoma cell lines, and expression of a constitutively active FOXO3 reduced tumorigenicity of glioblastoma cells[59]. Additionally, cytoplasmic FOXO3a and expression of IKKβ have been reported to associate with poor survival in breast cancer patients. IKKβ- stimulated tumorigenesis can be reversed by re-expression of FOXO3[51]. The preponderance of experimental evidence demonstrates that FOXOs function as efficient tumor suppressors when activated.

The anti-tumor activity of FOXOs has been linked to their capacity to transcriptionally activate several proapoptotic targets, including (FasL), TNF-related apoptosis inducing ligand (TRAIL), and

Bcl-2-interacting mediator of cell death (BIM). FOXO proteins are also able to promote cell cycle arrest via upregulation of P27, P21 and Cyclin G2, and repression of Cyclin D[60]. In spite of the well documented tumor suppressor function, recently several findings have demonstrated that FOXOs can also harbor pro-tumor effects. It was shown that FOXO3 remained nuclear in human anaplastic

72 carcinoma cells and supported cell cycle progression and proliferation through transcriptional upregulation of Cyclin A1[61]. Isocitrate dehydrogenases 1(IDH1) was reported as a target of FOXO1 and

FOXO3. FOXOs can activate mutant IDH1 expression as well, and thus maintain the levels of the oncometabolite 2-hydroxyglutarate, which stimulates proliferation of cancer cells harboring IDH1 mutation[62]. Additionally, colon cancer cells with high expression of nuclear β-Catenin somehow tolerated FOXO3-induced apoptosis, and simultaneous nuclear accumulation of β-Catenin and FOXO3 promoted metastasis instead[63]. Therefore, FOXOs can display either an inhibitory or a supportive role for tumorigenesis depending on the cell context.

FOXOs and stem cell maintenance

FOXO proteins are well known defenders against oxidative stress. In worms, flies and mice, genetic alterations that attenuate /IGF signaling which activate AKT or ERK downstream, with increased activity of FOXO resulting in extended lifespan[64-66]. This phenotype associates with enhanced resistance to oxidative stress which is considered as a determinant of ageing[67]. FOXOs regulate oxidative stress in part by upregulating the expression of several anti-oxidant that facilitate clearance of excessive reactive oxidative species (ROS), including superoxide dismutases (SOD) and catalase[68, 69]. DNA damage response gene GADD45a has been identified as an activating target of

FOXO3, which could promote DNA repair and thus the survival of the cells under stress conditions[70].

The protective function of FOXOs is critical for somatic stem cell maintenance. Conditional deletion of

Foxo1, 3 and 4 in mice caused a profound self-renewal defect in HSC and neural stem cells (NSC). Loss of quiescence and hyper-proliferation were found in these Foxos- deficient stem cells, which led to stem cell exhaustion[71, 72]. It seems that Foxo3 is the main FOXO regulating homeostasis of HSC and NSC in mice, as knockout of only Foxo3 was able to induce a similar defective phenotype[73-75]. In contrast,

Foxo1-/- or Foxo4-/- did not elicit any HSC defects[71], although their roles in NSC remain to be examined. Accumulation of ROS contributes to the loss of Foxo-/- stem cells, as it has been shown that reducing ROS by antioxidant N-acetyl L-cysteine (NAC) can partially rescue the self-renewal defects[71-

73

73]. In addition to regulating ROS, FOXOs can protect HSC and NSC via inducing autophagy, maintaining mitochondria function and coordinating to create a cellular environment benefiting stem cells[76-78]. FOXO1 and FOXO4 are also essential regulators of human embryonic stem cells (ESC), but exerting their function through distinct mechanisms. FOXO1 is required for activating pluripotency genes OCT4 and [79]. Instead it was reported that FOXO4 activates PSMD11 to increase proteasome activity, and the proteostasis resulting from a high proteasome activity is critical for pluripotency of ESC[80]. Additionally, FOXO1 appears to be crucial in mouse spermatogonial stem cells

(SSC) for their maintenance, with several key genes involved in self-renewal of SSC suggested as targets of FOXO1[81]. Thus, FOXO family members play extensive roles in stem cell maintenance by various mechanisms, not only ROS regulation.

FOXOs in leukemia

Given the complex biology of FOXOs, it is not surprising that FOXOs have a controversial role in leukemia. On one hand, many studies have shown they act as bona fide tumor suppressors. BCR-ABL is a constitutively active tyrosine kinase generated by t(9;22)(q34;q11), associating with both ALL and chronic myeloid leukemia (CML). BCR-ABL has been shown to inhibit FOXO3 activity via promotion of persistent AKT-dependent phosphorylation. FOXO3 inhibition is important for suppression of TRAIL- induced apoptosis and ID1-induced differentiation of BCR-ABL transformed cells[82, 83]. FOXO3 also acts as a key regulator of apoptosis induced by imatinib, a tyrosine kinase inhibitor targeting BCR-

ABL[84]. Moreover, promoting FOXO3 activity elicited apoptosis in BCR-ABL transformed murine B progenitors[85]. In AML cells, constitutively active IKK phosphorylates FOXO3 and results in its cytoplasmic localization, and restoring FOXO3 activity impaired cell growth[86]. A study of 511 patients showed that high levels of phosphorylation of FOXO3 is an independent adverse prognostic factor in

AML. Levels of phosphorylated FOXO3 were not associated with karyotype but were higher in patients with FLT3 mutations[87]. Accordingly, it was demonstrated that FLT3-ITD mutant protein inhibits

FOXO3 and it’s apoptosis induction by promoting PI3K dependent phosphorylation[88]. In addition,

74

FOXO3 was shown to localize in the cytoplasm in PML-RARα expressing AML cells, and became activated during all-trans retinoic acid (ATRA) treatment. Knockdown of FOXO3 led to inhibition of

ATRA-induced granulocytic differentiation or apoptosis[89].

On the other hand, in line with its critical role in HSC maintenance, FOXO3 was found to be important for maintaining leukemia stem cells (LSC). It was reported that FOXO3 is nuclear in LSC in a mouse

BCR-ABL CML model. Using serial transplantation, researchers showed that loss of Foxo3 increased survival of tertiary recipient mice, suggesting that Foxo3 deficiency led to LSC loss[90]. A later study identified BCL-6 as a FOXO3 target gene that represses p53, enhancing the survival and self-renewal of

CML LSC. A similar role of FOXOs was described in LSC of mouse MLL-AF9 AML. FOXO3 localized to the nucleus of AML LSC, and depletion of FOXOs promoted myeloid differentiation and cell death, leading to a reduction of LSC frequency[91]. However, the molecular mechanism by which FOXO3 is required for AML LSC was not defined, and it is therefore unclear whether FOXOs serve as stem cell maintenance genes, similar to their role in normal HSC, or as oncogenes, in that they can actively transform normal cells.

As mentioned above, different FOXO family members could have non-redundant functions [46]. Since most previous leukemia studies focused on FOXO3, the role of FOXO1 in pre-leukemia stem cells has not been explored. Here, we show that FOXO1 is not only required for the growth of AE pre-leukemia cells, it can function as an oncogene whose upregulation promotes self-renewal and blocks the differentiation of human CD34+ HSPCs. In t(8;21) AML, this oncogenic function is required for the activation of a self-renewal program. Our results demonstrate a new function of FOXO1 in AML pathogenesis, strengthening our understanding of the mechanisms that mediate the aberrant self-renewal of pre-leukemia stem cells and revealing potential therapeutic strategies for their elimination.

75

3.2 Results

FOXO1 is upregulated in AE leukemia and pre-leukemia cells

To identify critical regulators of self-renewal in AE AML, we examined an AML patient dataset [92] for stem cell related genes that are significantly upregulated in t(8;21) AML cells compared to other AML subtypes. FOXO1 was one such gene (Figures 3-1A and B). We validated the association between increased FOXO1 and AE AML in other datasets (Figure 3-1 C), and confirmed the upregulation of

FOXO1 protein in AE AML patient samples (Figure 3-1 D).

Figure 3-1. t(8;21) AML is associated with increased FOXO1. (A and B) Microarray analysis of FOXO1 transcript levels in (8;21) AML compared to non-t(8;21) AML (A, two probesets are shown) or to different cytogenetic groups (B). Dataset from Ross et al., 2004. *p=0.0042, **p=6.6×10-6, ***p=0.00014, p values were calculated by two-tailed t-test. (C) FOXO1 expression in t(8;21) AML compared to non-t(8;21) AML, from datasets GSE6891 and GSE17855. p values were calculated by two-tailed t-test. (D) Immunoblot showing FOXO1 protein levels in AML patient samples.

76

Increased transcript and protein levels were also evident in AE pre-leukemia cells compared to control vector transduced CD34+ HSPCs (Figures 3-2 A and B). We developed an AE-Tet-off system to test whether FOXO1 is a downstream target of AE, and found that the expression of FOXO1 was highly dependent on continued expression of AE (Figures 3-2 C and D). Accordingly, chromatin immunoprecipitation-sequencing (ChIP-seq) analysis in the t(8;21) patient cell line Kasumi-1 showed AE bound at the FOXO1 gene locus (Figure 3-2 E). We validated this binding using the anti-HA antibody to

ChIP the epitope-tagged AE protein in AE cells (Figure 3-2 F). These results suggest that FOXO1 is upregulated in AE cells at the transcriptional level.

77

Figure 3-2. FOXO1 is upregulated in AE pre-leukemia cells. (A and B) FOXO1 mRNA determined by qPCR(A, n=4, error bars represent SD) and protein(B) levels in AE- or empty vector (MIT) transduced CD34+ HSPCs. (C and D) AE tet-off system showed FOXO1 mRNA(C, n=3, error bars represent SD) and protein(D) decreased following turning off AE expression by DOX. (E) ChIP-seq showed AE bound to intron 1 of FOXO1 gene in Kasumi-1 cells. The sequence of binding sites is shown with AML1 binding motif highlighted. (F) ChIP-qPCR confirmed the AE (HA-tagged) binding on FOXO1 locus in AE cells, with primers indicated in (E) by arrow. RPL30 is used as negative control. n= 3, error bars represent SD. (G and H) Cellular fractionation immunoblot (G) and immunofluorescence analysis (H) showed FOXO1 mainly located in nucleus of AE cells. Scale bar=20µm.

The transcriptional activity of FOXOs is dependent on their nuclear localization[86]. Cellular fractionation and immunoblot analysis showed that FOXO1 has a primarily nuclear distribution in AE cells (Figure 3-2 G). Immunostaining confirmed this finding, suggesting that FOXO1 is active in AE pre- leukemia cells (Figure 3-2 H).

FOXO1 is required to sustain the growth of AE cells

To investigate whether FOXO1 functionally contributes to AE cell growth, we knocked down (KD)

FOXO1 using two independent shRNAs. FOXO1 KD impaired growth in liquid culture and clonogenicity of AE cells in a methylcellulose assay (Figures 3-3 A and B). Additionally, FOXO1-depleted AE cells showed a reduced long-term engraftment capacity in immunodeficient mice, as indicated by the reduction of the Venus+ fraction in engrafted human cells (Figure 3-3 C). These results suggest that FOXO1 does not act as a tumor suppressor, instead, FOXO1 ablation impairs the self-renewal of AE pre-leukemia cells both in vitro and in vivo. We also evaluated FOXO1’s function using a FOXO-specific inhibitor [93].

Inhibitor treatment dramatically reduced the growth and clonogenic potential of AE cells while showing only minimal effects on normal human CD34+ HSPCs (Figures 3-3 D and E). These findings are in accordance with data showing that Foxo1 knockout mice do not present with hematopoiesis defects [71], demonstrating that FOXO1 is specifically required in AE pre-leukemia cells.

78

Figure 3-3. FOXO1 inhibition impairs growth of AE cells. (A) AE cells were transduced with two sets of non- targeting shRNA and FOXO1 shRNA vectors that co-expressed Venus. Knockdown was confirmed by immunoblotting (upper). Growth of AE cells in liquid culture was measured as change of percentage of Venus+ cells relative to shNT by flow cytometry. (B) CFU assay of sorted shRNA-transduced AE cells. n=3, results represent mean +- SD. (C) shRNA-transduced AE cells were transplanted into immunodeficient mice. Venus+ percentage of human engrafted cells in bone marrow was examined 9 weeks later, and normalized to those before the transplantation. (*p<0.005, **p<0.001, two-tailed t-test). One representative experiment of 2 replicates is shown. (D) CD34+ HSPCs and AE cells were treated with the FOXO1 inhibitor AS1842856 or DMSO, cell numbers were monitored over time. One representative experiment of 2 replicates is shown. (E) CFU assay of CD34+ HSPCs and AE cells treated with FOXO1 inhibitor. n=3, results represent mean +- SD.

Increased FOXO1 has oncogenic activity in human CD34+ cells

The increase in FOXO1 upon AE expression in normal CD34+HSPCs suggested that FOXO1 could play a role in AE pathogenesis. To test this, we transduced CD34+ HSPCs with a retrovirus expressing wildtype FOXO1 (FOXO1 WT) or only GFP (MIG) as control. Since FOXO1 could have transcription independent functions [94], we also transduced a FOXO1 mutant unable to bind DNA (FOXO1 DB).

79

FOXO1 WT promoted the long-term proliferation of HSPCs in liquid culture, while control and FOXO1

DB cells grew for only a limited period (Figure 3-4 A). Immunophenotypic and morphologic analyses showed FOXO1 WT cells retained an immature morphology with very little CD11b expression while control and FOXO DB cells underwent terminal myeloid differentiation (Figures 3-4 B and C). In methylcellulose clonogenic assays, FOXO1 WT cells gave rise to significantly more myeloid colonies with a dramatic loss of erythroid colonies when compared to MIG and FOXO1 DB cells (Figure 3-4 D).

In addition, FOXO1 WT cells generated substantially more colonies in both secondary and tertiary re- platings, in striking contrast to control cells, where re-plating activity was successively diminished

(Figure 3-4 E). All these effects are similar to the phenotype elicited upon AE expression [95].

To investigate FOXO1’s effect in vivo, we transplanted transduced HSPCs into immunodeficient mice.

Twelve weeks after transplantation, the GFP+ percentage of engrafted human cells was measured.

Consistent with our previous observation [96], control MIG cells did not maintain long-term engraftment in vivo and thus the percentage of engrafted human cells expressing GFP decreased over time. FOXO DB cells showed a similar loss in vivo. However, for FOXO1 WT cells, the fraction of transduced cells was maintained (Figure 3-4 F). Nevertheless, as seen for AE cells, FOXO1 WT cells were unable to initiate

AML. Taken together, these data indicate that enforced expression of FOXO1 in normal CD34+ HSPCs enhances stem cell function while disrupting differentiation, partially phenocopying AE pre-leukemia cells and highlighting an oncogenic function of FOXO1.

To test whether this oncogenic function is a common feature for other FOXO proteins, FOXO3 was overexpressed in CD34+HSPCs as a comparison. However, unlike FOXO1, increasing FOXO3 expression did not elicit a growth advantage in liquid culture compared to MIG transduced cells (Figure

3-5 A). In methylcellulose clonogenic assays, FOXO3 did not promote myeloid colony formation, instead, FOXO3 cells tended to generate fewer colonies of both myeloid and erythroid lineages compared to MIG cells (Figure 3-5 B). Additionally, FOXO3 cells did not display enhanced replating ability as

80

FOXO1 cells did (Figure 3-5 C). Therefore, these data indicate that the oncogenic function is likely specific to FOXO1.

81

Figure 3-4. Increased FOXO1 promotes a pre-leukemic state in human CD34+ HSPCs. (A) Weekly cell count of liquid culture of CD34+ HSPCs transduced with wildtype FOXO1 (WT), a FOXO1 mutant without a DNA binding domain (DB), or control (MIG) retroviral vectors. FOXO1 expression was confirmed (upper). (B) Flow cytometry analysis of the myeloid differentiation marker CD11b on cells from a week 4 culture. (C) Wright-Giemsa staining of week 4 cultured cells. Scale bar=20 µm. (D and E) CFU assay with transduced cells. Results represent mean +- SD of colony counts of first round (D) and second and third rounds (E), n=3. (F) Transduced cells were transplanted into immunodeficient mice. GFP+ percentage of human engrafted cells in bone marrow was examined 12 weeks later, and normalized to those before the transplantation. (*p=0.019, **p<1×10-5, two-tailed t-test). One representative experiment of 2 replicates is shown.

Figure 3-5. FOXO3 does not display the oncogenic function. (A) Weekly cell count of liquid culture of CD34+ HSPCs transduced with FOXO1, FOXO3 or control (MIG) retroviral vectors. (B and C) CFU assay with transduced cells. Results represent mean +- SD of colony counts of first round (B) and second round (C), n=3.

82

The oncogenic function of FOXO1 is independent of ROS regulation

The restraint of ROS by FOXOs is required for HSC maintenance, and accordingly, use of the antioxidant

N-acetyl L-cysteine (NAC) can rescue HSC defects caused by FOXO deletion [71]. To investigate whether ROS regulation contributes to FOXO1’s oncogenic function, we first examined the intracellular

ROS level in transduced cells. Increased FOXO1 resulted in a modest reduction of intracellular ROS immediately after transduction, however, this effect was transient, as ROS levels in FOXO1 WT cells were comparable to control cells after two weeks (Figure 3-6 A). To test the biological influence of ROS reduction, we treated normal CD34+ HSPCs with NAC, which effectively decreased ROS levels (Figure

3-6 B, left). In contrast to the block of differentiation caused by FOXO1, NAC treatment only delayed the differentiation process, as CD11b expression on NAC-treated cells increased over time, although with slower kinetics than control cells (Figure 3-6 B, right). In striking contrast to the effects of FOXO1 which promoted long-term growth of CD34+ HSPCs, NAC treatment hampered cell proliferation rather than enhanced it, suggesting ROS regulation does not play a major role in facilitating the oncogenic function of FOXO1 (Figure 3-6 C).

83

Figure 3-6. ROS reduction cannot recapitulate the effects of increased FOXO1. (A) Flow cytometry analysis of intracellular ROS in CD34+ HSPCs transduced with MIG, FOXO1 WT or FOXO1 DB at indicated time points after transduction. The numbers represent median fluorescent intensity. (B) Flow cytometry analysis of intracellular ROS in CD34+ HSPCs treated with 2 mM NAC or ethanol as control (left). Expression of CD11b of treated cells was analyzed at indicated time points (right). (C) Weekly count of cell number of CD34+ HSPCs treated with 2mM NAC or ethanol. One representative of three independent experiments was shown.

Increased FOXO1 elicits a gene expression signature shared with AE

The transcriptional targets of FOXOs in normal HSCs and leukemia stem cells are poorly explored. To gain a better understanding of the underlying gene network accounting for FOXO1’s oncogenic role, we performed RNA-seq analysis on CD34+ HSPCs transduced with AE, FOXO1 WT, FOXO1 DB or MIG on day 5 post transduction. Hierarchical clustering analysis showed that gene expression patterns from

FOXO1 DB and MIG cells clustered together, indicating that FOXO1 DB does not have a major impact on gene expression and thus is non-functional in this context. In contrast, gene expression patterns from

AE and FOXO1 WT cells clustered separately from the control cells. More strikingly, a significant portion of the AE gene expression signature was also present in FOXO1 cells (Figure 3-7 A).

Accordingly, pathway enrichment analysis on genes activated by FOXO1 compared to control cells showed a significant enrichment of published gene signatures characterizing AE pre-leukemia and leukemia. In addition, HSC signature genes were also significantly enriched (Figure 3-7 B). These results suggest that FOXO1 regulates a core network associating with HSC function that is critical for AE mediated pre-leukemia stem cell programming.

84

Figure 3-7. Transcriptome of FOXO1 recapitulates gene signature of AE. (A) Heat map depicting hierarchical clustering of significantly differentially expressed genes of AE cells compared to MIG cells. (B) Pathway enrichment analysis of FOXO1 WT target genes showed enrichment of AE and HSC signatures.

FOXO1 and AE show genome-wide co-localization and FOXO1 maintains activation of key AE target genes

To identify the genomic targets of FOXO1 we performed ChIP-Seq analysis of AE and FOXO1 in AE pre-leukemia cells. In total, 14723 AE- and 11835 FOXO1-bound loci were identified. Peak distributions differed slightly, with FOXO1 showing a higher localization to promoter regions than AE (Figure 3-8 A).

In agreement with our previous report showing FOXO binding motif enrichment in AE-bound loci

85 mapped in established leukemic cells [35], we found 61% of AE and 76% of FOXO1 bound loci were in common (Figure 3-8 B). Unbiased motif enrichment analysis in joint peaks showed that both RUNX1 and

FOXO motifs were significantly enriched, together with ERG/ETS and EBox motifs, consistent with our previous report (Figures 3-8 C and D) [35]. Taken together, these results suggest that the FOXO1 and AE molecular networks are widely interconnected.

86

Figure 3-8. FOXO1 regulates AE key target genes. (A) Genomic location of AE and FOXO1 binding sites. (B) Venn diagram showing the overlap between AE and FOXO1 ChIP-seq peaks. (C) Enriched transcription factor binding motifs in jointly bound AE and FOXO1 ChIP-seq peaks. (D) Percentage of peaks containing predicted enriched motifs. (E) Percentage of AE regulated genes associated with both AE and FOXO1 binding. (F) UCSC browser screenshot showing AE and FOXO1 binding sites on SOX4, MPL and UBASH3B loci. Arrows indicate jointly bound AE and FOXO1 peaks. (G) Immunoblot showing decreased protein levels of SOX4, MPL and UBASH3B protein upon FOXO1 knockdown in AE cells.

Cross analysis with RNA-seq data showed that about 50% of AE regulated genes have both AE and

FOXO1 bound in close proximity to the gene body (Figure 3-8 E), including MPL, UBASH3B and SOX4

(Figure 3-8 F). We and others have reported that AE leukemia and pre-leukemia cells depend on MPL and UBASH3B for self-renewal and long-term proliferation [42, 43, 96]. In addition, SOX4 is a critical oncogenic mediator of CEBPA mutant AML and efficiently transforms mouse HSPCs [97]. Interestingly,

FOXO1 KD in AE cells led to the downregulation of all three genes (Figure 3-8 G), which may account for the impaired growth of AE cells. These data suggest that the activation of a substantial set of AE targets cannot be fulfilled by AE itself and requires the up-regulation of FOXO1, thus facilitating pre- leukemia stem cell programming.

3.3 Discussion

Understanding the mechanisms through which leukemia oncogenes reprogram transcriptional networks and lead to enhanced self-renewal will provide important insights in our efforts to target leukemic and pre-leukemic stem cells and thus prevent relapse. The key oncogenic mediators for leukemic programming of several subtypes of AML have been identified, including HOXA9/MEIS1 for MLL- translocation related AML and SOX4 for AML associated with CEBPA mutations [97, 98]. Yet how AE promotes self-renewal and whether a key mediator exists in t(8;21) AML has been less clear. In this study, we identified FOXO1 as a critical regulator of the self-renewal program in AE pre-leukemia cells that is necessary and sufficient to initiate a pre-leukemic phenotype. We show FOXO1 is required for full activation of a set of self-renewal genes and the long-term proliferation and clonogenicity of AE cells.

87

Importantly, enforced expression of FOXO1 in normal CD34+ HSPCs can partially recapitulate the cellular phenotype and gene signature of AE cells, revealing increased FOXO1 as a new molecular mechanism by which pre-leukemia stem cells establish an aberrant self-renewal program. Previous studies have revealed that WNT/β-Catenin pathway plays an important role in self-renewal maintenance of AE leukemia stem cells[37]. Of note, it was reported that β -Catenin interacts with FOXO and enhanced its activity in regulating oxidative stress[99]. It would be interesting to investigate whether

WNT signaling goes through FOXO1 to exert self-renewal promotion in AE pre-leukemia stem cells.

We showed that FOXO1 localized to the nucleus in AE cells, and its function depended on its transcription activity. As reported by us and others, PI3K/AKT signaling is activated in AE leukemia cells[42, 43], therefore, other mechanisms exist to facilitate nuclear transportation of FOXO1 regardless of AKT activation. Several post-translational modifications of FOXOs can enhance FOXOs activity and override repression by AKT. JNK can phosphorylate FOXOs and promote their import into the nucleus[53]. Activation of JNK signaling by AE has been reported[100]. In addition, methylation of

FOXO1 by protein methyltransferase PRMT1 inhibits the AKT-mediated phosphorylation[101].

Intriguingly, PRMT1 was shown to interact with AE to promote its transcription activity, which is required for AE cell proliferation[102]. It is possible that PRMT1 contributes to AE leukemogenesis through multiple aspects. Future studies demonstrating how FOXO1 activity is regulated in AE pre- leukemia cells will provide additional opportunities for blocking FOXO1 and ablating self-renewal of AE cells.

It was reported that FOXOs are required for maintaining AML stem cells [91], which might reflect the essential role of FOXOs in normal HSC maintenance. In contrast, whether FOXOs can act as an oncogene and actively transform normal cells to AML upon increased expression or activity is presently unknown.

Here, we demonstrated a clear gain-of function oncogenic role of FOXO1 in human HSPC, highlighting a new functional aspect of FOXOs in AML. Upregulation of FOXO1 is also seen in some non-t(8;21) AML

[103]. It is therefore possible that FOXO1 plays a more widespread role in AML development.

88

Interestingly, we found the oncogenic function was specific to FOXO1 while not displayed by FOXO3, emphasizing the non-redundant roles of FOXO proteins. It is possible that FOXO3 is more important in

AML maintenance but not initiation stages. Since the core DNA sequences of FOXOs binding sites are the same, the molecular mechanism mediating the specificity of different FOXO members is puzzling. It could be that they are involved in different protein-protein interactions or subject to distinct post- translational modifications. Also, it was reported that the different FOXO proteins showed distinct preference for the flanking DNA sequences around the core binding sites in vitro[44], which may influence their target gene selection in vivo.

In contrast to its essential role in HSC maintenance, ROS restriction is unlikely a major driver of the oncogenic phenotype upon increased FOXO1 in human HSPC. This is consistent with the finding that a physiological level of ROS is required for cell proliferation [104], and increased ROS favors AML progression [105]. Our data instead suggest that the oncogenic function of FOXO1 in AML is through the activation of a HSC program, demonstrating a cell context dependent function of FOXO1.

By virtue of selecting and regulating a set of AE target genes, FOXO1 reinforces the AE molecular network. How the specificity of FOXO1 binding to target genes is determined is not well understood. It was reported that FOXO3 binds to pre-existing enhancers in the genome[106]. The target gene preference of FOXOs can also be influenced through interaction with other transcription factors [107]. ETS transcription factors ERG and FLI1 have been reported to direct AE to specific binding regions [108].

Genomic co-localization of ETS factors and FOXO in non-hematopoietic tissues has been suggested

[109]. As evidenced by our finding that ERG and ETS motifs were significantly enriched in FOXO1 bound loci, it is possible ERG/FLI1 shape the genome landscape, and their binding sites also demarcate targets for FOXO1, thus leading to the co-selection of targets by FOXO1 and AE.

With the understanding that FOXOs are tumor suppressors, restoring FOXO activity is being considered as a potential cancer therapy [60]. Given the various functions of FOXOs in AML, caution has to be taken in applying this strategy until we fully understand the AML subtype specific role of FOXOs and evaluate

89 the possible oncogenic effects on normal HSPCs. Considering the selective toxicity of the FOXO inhibitor on AE cells, and the demonstration that an HSC defect was not seen in Foxo1- deleted mice[71],

FOXO1 may be an effective therapeutic target for specifically eliminating pre-leukemia stem cells in AE and other FOXO1-overexpressing AMLs.

3.4 Experimental Procedures

Plasmids and reagents

MSCV-IRES-GFP(MIG) and MSCV-AE-IRES-GFP retroviral vector was as described[95]. pSIN-

TREtight-DsRED-IRES(tTRi) and MSCV-GFP-IRES-tTA retroviral vectors were generously provided by Dr. Johannes Zuber. Insertion of AE was subcloned into tTRi to build an AE-tet-off system. pSG5L-

FOXO1-WT and pBabe- puro-FOXO1 DB vectors were obtained from Addgene (#10693 and #10695).

Insertions of FOXO1 WT and FOXO1 DB were then subcloned into MIG. Lentiviral vector MISSION pLKO.1-shRNA-puro constructs targeting human FOXO1 (TRCN0000039578, sh-1) and non-target shRNA control (NT-1) were obtained from Sigma, the puromycin-resistant gene in the constructs was replaced with Venus marker. GIPZ lentiviral shRNA targeting FOXO1 (V3LHS_405827, sh-2) and non- silencing control (NT-2) were obtained from Open Biosystems, the shRNA insertions were subcloned into SF-LV-shRNA-EGFP lentiviral vector, a gift from Dr. K. Lenhard Rudolph. FOXO1 inhibitor

AS1842856 was from Millipore (344355), dissolved in DMSO. NAC was from Sigma (A7250), dissolved in ethanol.

Cells and culture

Human umbilical cord blood cells (CB) were obtained by the Translational Trials Support Laboratory at

CCHMC under institution-approved protocol. CD34+ cells were enriched using CD34+ selection kit

(Miltenyi). AE pre-leukemia cells were established as described[30]. AE cells and MIG/FOXO1- transduced CB cells were cultured in IMDM with 20% BIT Serum Substitute (Stemcell Technologies

90

#9500), supplemented with 10 ng/mL SCF, IL-3, IL-6, FLT3L, and TPO. Patient samples were described in previous study[43].

Viral production and transduction

Retrovirus was produced in 293T cells transfected with retroviral vectors, together with envelope RD114 and the gag-pol M57 constructs. CD34+CB cells were stimulated IMDM with 20% BIT, SCF, FLT3L, and TPO (100 ng/mL) for 24 hours before transduction. Retronectin-coated plates (Takara) were preloaded with retroviral supernatant once by centrifuge at 2200 rpm and 4 °C for 30 min. Stimulated cells were then seeded into loaded plates with presence of additional retroviral supernatant on for 1 day.

CB cells were co-transduced with tTRi-AE and MSCV-tTA viruses in order to establish AE-tet-off systems. To produce lentivirus, 293T cells were transfected with lentiviral shRNA vectors with lentiviral envelope and gag-pol constructs. AE cells were mixed with lentiviral supernatant and transduced on retronectin-coated plates.

Flow cytometry and cell sorting

Antibodies (all BD unless noted) used were APC- and PE-CD33 (WM53), APC-CD11b (ICRF44), V450- human CD45 (HI30), APCCy7-mouse CD45 (30-F11). Cells from mouse tissues were incubated with nonspecific binding blocker (anti-mouse/human CD16/CD32 Fc g receptor; BD) before staining. To examine the level of intracellular ROS, cells were stained by using CellROX deep red reagent (Life

Technologies C10422). Cells were analyzed on FACSCanto flow cytometer (BD) or sorted on FACSAria

(BD) or MoFlo XDP (Beckman Coulter), and the data was analyzed with FloJo software (TreeStar).

Methylcellulose colony-forming assays

Assays were performed in MethoCult H4100 medium (Stemcell Technologies) supplemented with 20%

BIT, 50 µM -Mercaptoethanol, 2 mM L-glutamine, 100 U/mL penicillin/streptomycin, and the human cytokines Erythropoietin (6 U/mL), Granulocyte Colony Stimulating Factor (10 ng/mL), IL-6(20 ng/mL),

IL-3 (20 ng/mL), and SCF (20 ng/mL). Colonies were scored at day 14, and then the cells were collected

91 for replating. Some experiments were performed by using MethoCult Express medium (Stemcell

Technologies), colonies were scored at day 7.

Xenograft transplantation

Trasplantation was performed on NOD/SCID/IL2RG-/- immunodeficient mice with transgenic expression of human SCF, granulocyte–macrophage colony-stimulating factor and interleukin-3(NSGS). 6- to 8- week-old mice were conditioned with 30 mg/kg busulphan (Sigma) through intraperitoneal injection 24 hours before transplantation, 100×105 cells were transplanted through intrafemoral injection immediately after transduction. Bone marrow was aspirated for flow cytometry at indicated time. All experiments were performed in accordance with CCHMC institutional guidelines.

Immunostaining

Cytospin slides of AE cells were fixed in 4% paraformaldehyde, permeabilized with PBS containing 0.

5% Trition X-100, and blocked with 10% FBS in PBS. Then cells were incubated with primary antibody against FOXO1(Cell Signaling Technology #2880) at 4 °C overnight, and then with goat anti-rabbit IgG

(H+L) secondary antibody, Alexa Fluor 568 (Molecular Probes) for 1 hour at room temperature.

Western blotting

Nuclear lysates were obtained using NE-PER nuclear extraction kit (Thermo Scientific). Protein samples were run on 6% or 8% polyacrylamide gels and transferred to a nitrocellulose membrane. The primary antibodies used were anti-FOXO1 (Cell Signaling Technology #2880), anti-MPL (Millipore 06944), anti-

UBASH3B (Abcam ab34781) , anti-SOX4 (Sigma AV38234) , anti-HDAC1 (Cell Signaling Technology

#5363), anti-ACTIN (Sigma A3854) and anti-TUBULIN (Sigma T0198). Anti-ETO antibody was a gift from Dr.Inge Olsson. The secondary antibodies were HRP-linked goat anti-rabbit IgG and anti-mouse

IgG at 1:1000 (Cell Signaling Technology #7074, #7076)

92

RT-PCR

RNA was reversed transcribed using MuLV Reverse Transcriptase and random hexamers (Applied

Biosystems). The cDNA was then subject to qPCR using SYBR Green technology (Roche). Expression level was calculated by ∆∆Ct method, normalized to PGK1. Primers are:

PGK1 F GGGAAAAGATGCTTCTGGGAA; PGK1 R TTGGAAAGTGAAGCTCGGAAA;

FOXO1 F TCGGCGGGCTGGAAGAATTCAA; FOXO1 R TTTCCCGCTCTTGCCACCCTC;

AE F CACCTACCACAGAGCCATCAAA; AE R ATCCACAGGTGAGTCTGGCATT.

RNA sequencing

Three different units of CD34+ CB cells were transduced with MIG, AE, FOXO1 WT and FOXO1 DB.

RNA was collected from sorted GFP+CD11b- cells 5 days after transduction using RNeasy Mini Kit

(Qiagen).The integrity of RNA was analyzed by Bioanalyzer (Agilent). 1 µg total RNA was used for poly(A)

RNA selection, followed by cDNA synthesis using PrepX mRNA Library kit (WaferGen) and Apollo 324

NGS automatic library prep system. Sample-specific index was added to the adaptor-ligated cDNA by PCR with index-specific primers for 13 cycles. The cluster generation of indexed libraries was carried out on cBot system (Illumina) using Illumina’s TruSeq SR Cluster kit v3, and then sequenced on Illumina HiSeq system using TruSeq SBS kit to generate single-end 50 cycle reads. 20-50 million reads were generated for each sample.

Analysis of RNA sequencing data

RNAseq data was processed with Tophat/Cufflinks V using the UCSC ref 38 gene/transcript model.

FPKM values less than 0.1 were adjusted to 0.1, and normalized per sample to the 60th %ile per sample, and then to the median of the 3 MIT empty vector treated samples or to 1, whichever was higher.

Differentially expressed genes were identified by filtering for genes that were expressed at a level greater than 2 FPKM in at least two samples and that differed between MIT, AE, FOXO1-WT, and FOXO1-DB

93 with p<0.05 using students t-test ANOVA. The normalized expression values were then subjected to hierarchical clustering using Pearson Correlation.

Pathway enrichment analysis

The pathway enrichment analysis was performed for FOXO1 WT-upregulated genes using the LRpath methodology with the gene lists from the MSigDB database[110].

Chromatin immunoprecipitation

ChIP-Seq assays in AE cells were performed essentially as previously described[35]. Antibodies for HA

(Sigma, H6908) and FOXO1 (Abcam ab39670) were used. Cells were first washed with PBS and then cross-linked with 0.83 mg/ml Di(N-succinimidyl) glutarate (DSG) (Sigma) for 45 minutes at room temperature with rotation. Cells were washed four times with PBS and further crosslinked in PBS with

1% formaldehyde (~0.34 M) for 10 min at room temperature. All crosslinking reactions were quenched by adding 4 volumes of PBS and 0.125 M glycine, cells were washed twice in PBS and incubated with

Buffer A (10 mM HEPES pH 8.0, 10 mM EDTA, 0.5 mM EGTA, 0.25% Triton X-100, proteinase inhibitor cocktail (Roche UK, Burgess Hill, UK) and 0.1 mM PMSF), incubated for 10 min at 4 °C with rotation, and centrifuged 5 min at 500 g at 4 °C. The pellet was resuspended in 10 ml of ice–cold Buffer B

(10 mM HEPES pH 8.0, 200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.01% Triton X-100, protease inhibitor cocktail and 0.1 mM PMSF), incubated for 10 min at 4 °C with rotation and centrifuged for 5 min at 500 g at 4 °C. Cells were resuspended in 600 µl of ice-cold ChIP lysis buffer (25 mM Tris-HCl pH

8.0, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, 0.25% SDS, protease inhibitor cocktail and 0.1 mM

PMSF), incubated 10 min on ice and sonicated at 5 °C using a Bioruptor™ (Diagenode) to generate fragments an average length of 500 bp (10 min with 30 s “ON” and “OFF” cycles, power setting high).

The lysates were centrifuged for 5 min at 16,000 g at 4 °C and the supernatants were diluted with two volumes of ice-cold ChIP dilution buffer (25 mM Tris-HCl pH 8.0, 150 mM NaCl, 2 mM EDTA, 1%

Triton X-100, 7.5% glycerol, protease inhibitor cocktail and 0.1 mM PMSF). For each IP, 15 µl of

94

Dynabeads® protein G (Dynal) were pre–incubated with 50 µg BSA and 1-2 µg antibody for 2 hours at

4 °C with rotation. The blocked antibody-bound protein G mix was added to 20–25 µg chromatin in a total volume of 500 µl diluted ChIP lysis buffer and incubated for 2 hours at 4°C with rotation. After magnetic separation the beads were washed once with 1 ml wash buffer 1 (20 mM Tris-HCl pH 8.0, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, 0.1% SDS), twice with 1 ml wash buffer 2 (20 mM Tris-HCl pH 8.0, 500 mM NaCl, 2 mM EDTA, 1% Triton X-100, 0.1% SDS), once with 1 ml LiCl buffer (10 mM

Tris-HCl pH 8.0, 250 mM LiCl, 1 mM EDTA, 0.5% NP-40, 0.5% Na-deoxycholate) and twice with 1 ml

TE/NaCl buffer (10 mM Tris-HCl pH 8.0, 50 mM NaCl, 1 mM EDTA). For each wash the beads were mixed with ice-cold washing buffers for 10 min at 4 °C. The immunoprecipitated DNA was eluted two times with 50 µl ChIP elution buffer (100 mM NaHCO3, 1% SDS) for 15 min at RT with shaking. At this step the input control (1% of the starting material) was included in the experimental procedure after first adjusting the final volume to 100 µl with ChIP elution buffer. The eluted DNA was incubated overnight at 65 °C in the presence of 50 µg proteinase K. The DNA was finally purified using Agencourt AMPure

(Beckman Coulter) magnetic beads according to the manufacturer’s instructions, eluted with 50 µl TE.

ChIP library preparation

DNA libraries for sequencing were prepared from approximately 10 ng DNA from ChIP samples using the KAPA Hyper (KR096100) library preparation kit according to the manufacturer’s instructions (Kappa

Biosystems).

ChIP data analysis

-Alignment

Sequences reads in fastq format were mapped onto the reference human genome version hg38, Genome

Reference Consortium GRCh38. The Illumina reads were aligned to the human genome using

Bowtie2[111]. Reads that were uniquely aligned to chromosomal positions were retained and duplicate reads were removed from the aligned data using Picard tools (http://broadinstitute.github.io/picard/). The

95 filtered aligned reads were used to generate density profiles using “genomeCoverageBed” function from bedtools (http://bedtools.readthedocs.org/en/latest/). These tag densities were displayed using the UCSC

Genome Browser.

-Peak calling

Regions of enrichment (peaks) of ChIP sequencing data were identified using DFilter software[112] with recommended parameters (-bs=100 -ks=50 –refine). Peak overlaps, gene annotations were performed using in-house scripts. Peaks were allocated to genes if located in either their promoters or within the region of 500 bp downstream and 2000 bp upstream of the transcription start sites (TSS), as intragenic if not in the promoter but within the gene body region, or if intergenic, to the nearest gene located within

100 kb. Overlaps between ChIP-seq peaks were defined by requiring the summits of two peaks to lie within +/-200 pb.

-Motif analysis

De novo motif analysis was performed on peaks using HOMER[113]. Motif lengths of 6, 8, 10, and 12 bp were identified in within ± 200 bp from the peak summit and a random background sequence option was used. The motif matrices generated by HOMER were scanned against JASPAR with the use of STAMP to identify similarity to known transcription factor binding sites. The top enriched motifs with a significant p value score were recorded. The annotatePeaks function in HOMER was used to find occurrences of motifs in peaks. In this case we used known motif position weight matrices (PWM) from

HOMER database. Percentages of motifs are displayed as a heatmap.

-Correlation of gene expression and ChIP-seq data

Genes with at least two fold-changes in expression (either up or down) that changed expression in AE versus MIG were selected and correlated with FOXO1 and AE ChIP-Seq co-localized genes. p values were calculated by hypergeometric test using phyper function in R.

96

3.5 Reference

1. Muller, A.M., et al., Complementing mutations in core binding factor leukemias: from mouse models to clinical applications. Oncogene, 2008. 27(44): p. 5759-73. 2. Bitter, M.A., et al., Associations between morphology, karyotype, and clinical features in myeloid leukemias. Hum Pathol, 1987. 18(3): p. 211-25. 3. Miyoshi, H., et al., The t(8;21) translocation in acute myeloid leukemia results in production of an AML1-MTG8 fusion transcript. EMBO J, 1993. 12(7): p. 2715-21. 4. Miyoshi, H., et al., t(8;21) breakpoints on chromosome 21 in acute myeloid leukemia are clustered within a limited region of a single gene, AML1. Proc Natl Acad Sci U S A, 1991. 88(23): p. 10431-4. 5. Erickson, P., et al., Identification of breakpoints in t(8;21) acute myelogenous leukemia and isolation of a fusion transcript, AML1/ETO, with similarity to Drosophila segmentation gene, runt. Blood, 1992. 80(7): p. 1825-31. 6. Nisson, P.E., P.C. Watkins, and N. Sacchi, Transcriptionally active chimeric gene derived from the fusion of the AML1 gene and a novel gene on in t(8;21) leukemic cells. Cancer Genet Cytogenet, 1992. 63(2): p. 81-8. 7. Okuda, T., et al., AML1, the target of multiple chromosomal translocations in human leukemia, is essential for normal fetal liver hematopoiesis. Cell, 1996. 84(2): p. 321-30. 8. Wang, Q., et al., Disruption of the Cbfa2 gene causes necrosis and hemorrhaging in the central nervous system and blocks definitive hematopoiesis. Proceedings of the National Academy of Sciences of the United States of America, 1996. 93(8): p. 3444-9. 9. Wang, Q., et al., The CBFbeta subunit is essential for CBFalpha2 (AML1) function in vivo. Cell, 1996. 87(4): p. 697-708. 10. Sasaki, K., et al., Absence of fetal liver hematopoiesis in mice deficient in transcriptional coactivator core binding factor beta. Proceedings of the National Academy of Sciences of the United States of America, 1996. 93(22): p. 12359-63. 11. Chen, M.J., et al., Runx1 is required for the endothelial to haematopoietic cell transition but not thereafter. Nature, 2009. 457(7231): p. 887-91. 12. Cai, X., et al., Runx1 loss minimally impacts long-term hematopoietic stem cells. PLoS One, 2011. 6(12): p. e28430. 13. Growney, J.D., et al., Loss of Runx1 perturbs adult hematopoiesis and is associated with a myeloproliferative phenotype. Blood, 2005. 106(2): p. 494-504. 14. Davis, J.N., L. McGhee, and S. Meyers, The ETO (MTG8) gene family. Gene, 2003. 303: p. 1-10. 15. Okuda, T., et al., Expression of a knocked-in AML1-ETO leukemia gene inhibits the establishment of normal definitive hematopoiesis and directly generates dysplastic hematopoietic progenitors. Blood, 1998. 91(9): p. 3134-43. 16. Yergeau, D.A., et al., Embryonic lethality and impairment of haematopoiesis in mice heterozygous for an AML1-ETO fusion gene. Nature Genetics, 1997. 15(3): p. 303-6. 17. Castilla, L.H., et al., Failure of embryonic hematopoiesis and lethal hemorrhages in mouse embryos heterozygous for a knocked-in leukemia gene CBFB-MYH11. Cell, 1996. 87(4): p. 687-96. 18. Okumura, A.J., et al., t(8;21)(q22;q22) Fusion proteins preferentially bind to duplicated AML1/RUNX1 DNA-binding sequences to differentially regulate gene expression. Blood, 2008. 112(4): p. 1392-401. 19. Li, Y., et al., Genome-wide studies identify a novel interplay between AML1 and AML1/ETO in t(8;21) acute myeloid leukemia. Blood, 2016. 127(2): p. 233-42. 20. Ptasinska, A., et al., Depletion of RUNX1/ETO in t(8;21) AML cells leads to genome-wide changes in chromatin structure and transcription factor binding. Leukemia, 2012. 26(8): p. 1829-41.

97

21. Ben-Ami, O., et al., Addiction of t(8;21) and inv(16) acute myeloid leukemia to native RUNX1. Cell Rep, 2013. 4(6): p. 1131-43. 22. Shima, T., et al., The ordered acquisition of Class II and Class I mutations directs formation of human t(8;21) acute myelogenous leukemia stem cell. Exp Hematol, 2014. 42(11): p. 955-65 e1- 5. 23. Jan, M. and R. Majeti, Clonal evolution of acute leukemia genomes. Oncogene, 2013. 32(2): p. 135-40. 24. Ford, A.M., et al., Protracted dormancy of pre-leukemic stem cells. Leukemia, 2015. 29(11): p. 2202-7. 25. Shlush, L.I., et al., Identification of pre-leukaemic haematopoietic stem cells in acute leukaemia. Nature, 2014. 506(7488): p. 328-33. 26. Hospital, M.A., et al., Core-binding factor acute myeloid leukemia in first relapse: a retrospective study from the French AML Intergroup. Blood, 2014. 124(8): p. 1312-9. 27. Wiemels, J.L., et al., In utero origin of t(8;21) AML1-ETO translocations in childhood acute myeloid leukemia. Blood, 2002. 99(10): p. 3801-5. 28. Miyamoto, T., I.L. Weissman, and K. Akashi, AML1/ETO-expressing nonleukemic stem cells in acute myelogenous leukemia with 8;21 chromosomal translocation. Proc Natl Acad Sci U S A, 2000. 97(13): p. 7521-6. 29. Tsukamoto, N., et al., Recurrence of acute myelogenous leukemia with the same AML1/ETO breakpoint as at diagnosis after complete remission lasting 15 years: analysis of stored bone marrow smears. Int J Hematol, 2003. 78(4): p. 362-9. 30. Mulloy, J.C., et al., Maintaining the self-renewal and differentiation potential of human CD34+ hematopoietic cells using a single genetic element. Blood, 2003. 102(13): p. 4369-76. 31. Tonks, A., et al., Transcriptional dysregulation mediated by RUNX1-RUNX1T1 in normal human progenitor cells and in acute myeloid leukaemia. Leukemia, 2007. 21(12): p. 2495-505. 32. Pabst, T., et al., AML1-ETO downregulates the granulocytic differentiation factor C/EBPalpha in t(8;21) myeloid leukemia. Nat Med, 2001. 7(4): p. 444-51. 33. Vangala, R.K., et al., The myeloid master regulator transcription factor PU.1 is inactivated by AML1-ETO in t(8;21) myeloid leukemia. Blood, 2003. 101(1): p. 270-7. 34. Choi, Y., et al., Erythroid inhibition by the leukemic fusion AML1-ETO is associated with impaired acetylation of the major erythroid transcription factor GATA-1. Cancer Res, 2006. 66(6): p. 2990- 6. 35. Ptasinska, A., et al., Identification of a dynamic core transcriptional network in t(8;21) AML that regulates differentiation block and self-renewal. Cell Rep, 2014. 8(6): p. 1974-88. 36. Wang, Y., et al., The Wnt/beta-catenin pathway is required for the development of leukemia stem cells in AML. Science, 2010. 327(5973): p. 1650-3. 37. Zhang, Y., et al., AML1-ETO mediates hematopoietic self-renewal and leukemogenesis through a COX/beta-catenin signaling pathway. Blood, 2013. 121(24): p. 4906-16. 38. Zheng, X., et al., Gamma-catenin contributes to leukemogenesis induced by AML-associated translocation products by increasing the self-renewal of very primitive progenitor cells. Blood, 2004. 103(9): p. 3535-43. 39. Steffen, B., et al., AML1/ETO induces self-renewal in hematopoietic progenitor cells via the Groucho-related amino-terminal AES protein. Blood, 2011. 117(16): p. 4328-37. 40. Cheng, C.K., et al., Secreted-frizzled related protein 1 is a transcriptional repression target of the t(8;21) fusion protein in acute myeloid leukemia. Blood, 2011. 118(25): p. 6638-48. 41. Muller-Tidow, C., et al., Translocation Products in Acute Myeloid Leukemia Activate the Wnt Signaling Pathway in Hematopoietic Cells. Molecular and Cellular Biology, 2004. 24(7): p. 2890- 2904.

98

42. Pulikkan, J.A., et al., Thrombopoietin/MPL participates in initiating and maintaining RUNX1-ETO acute myeloid leukemia via PI3K/AKT signaling. Blood, 2012. 120(4): p. 868-79. 43. Chou, F.S., et al., The thrombopoietin/MPL/Bcl-xL pathway is essential for survival and self- renewal in human preleukemia induced by AML1-ETO. Blood, 2012. 120(4): p. 709-19. 44. Furuyama, T., et al., Identification of the differential distribution patterns of mRNAs and consensus binding sequences for mouse DAF-16 homologues. Biochem J, 2000. 349(Pt 2): p. 629- 34. 45. Jacobs, F.M., et al., FoxO6, a novel member of the FoxO class of transcription factors with distinct shuttling dynamics. J Biol Chem, 2003. 278(38): p. 35959-67. 46. Hosaka, T., et al., Disruption of forkhead transcription factor (FOXO) family members in mice reveals their functional diversification. Proc Natl Acad Sci U S A, 2004. 101(9): p. 2975-80. 47. Brunet, A., et al., Akt promotes cell survival by phosphorylating and inhibiting a Forkhead transcription factor. Cell, 1999. 96(6): p. 857-68. 48. Obsilova, V., et al., 14-3-3 Protein interacts with nuclear localization sequence of forkhead transcription factor FoxO4. Biochemistry, 2005. 44(34): p. 11608-17. 49. Huang, H., et al., Skp2 inhibits FOXO1 in tumor suppression through -mediated degradation. Proc Natl Acad Sci U S A, 2005. 102(5): p. 1649-54. 50. Brunet, A., et al., Protein kinase SGK mediates survival signals by phosphorylating the forkhead transcription factor FKHRL1 (FOXO3a). Mol Cell Biol, 2001. 21(3): p. 952-65. 51. Hu, M.C., et al., IkappaB kinase promotes tumorigenesis through inhibition of forkhead FOXO3a. Cell, 2004. 117(2): p. 225-37. 52. Yang, J.Y., et al., ERK promotes tumorigenesis by inhibiting FOXO3a via -mediated degradation. Nat Cell Biol, 2008. 10(2): p. 138-48. 53. Essers, M.A., et al., FOXO transcription factor activation by oxidative stress mediated by the small GTPase Ral and JNK. EMBO J, 2004. 23(24): p. 4802-12. 54. Zhang, X., et al., Regulation and function of FoxO transcription factors in normal and cancer stem cells: what have we learned? Curr Drug Targets, 2011. 12(9): p. 1267-83. 55. Paik, J.H., et al., FoxOs are lineage-restricted redundant tumor suppressors and regulate endothelial cell homeostasis. Cell, 2007. 128(2): p. 309-23. 56. Dong, X.Y., et al., FOXO1A is a candidate for the 13q14 tumor suppressor gene inhibiting signaling in prostate cancer. Cancer Res, 2006. 66(14): p. 6998-7006. 57. Gu, T.L., et al., NPM-ALK fusion kinase of anaplastic large-cell lymphoma regulates survival and proliferative signaling through modulation of FOXO3a. Blood, 2004. 103(12): p. 4622-9. 58. Ericson, K., et al., Genetic inactivation of AKT1, AKT2, and PDPK1 in human cells clarifies their roles in tumor growth regulation. Proc Natl Acad Sci U S A, 2010. 107(6): p. 2598- 603. 59. Sunayama, J., et al., FoxO3a functions as a key integrator of cellular signals that control glioblastoma stem-like cell differentiation and tumorigenicity. Stem Cells, 2011. 29(9): p. 1327- 37. 60. Singh, A., J. Plati, and R. Khosravi-Far, Harnessing the tumor suppressor function of FOXO as an alternative therapeutic approach in cancer. Curr Drug Targets, 2011. 12(9): p. 1311-21. 61. Marlow, L.A., et al., Foxo3a drives proliferation in anaplastic thyroid carcinoma through transcriptional regulation of cyclin A1: a paradigm shift that impacts current therapeutic strategies. J Cell Sci, 2012. 125(Pt 18): p. 4253-63. 62. Charitou, P., et al., FOXOs support the metabolic requirements of normal and tumor cells by promoting IDH1 expression. EMBO Rep, 2015. 16(4): p. 456-66. 63. Tenbaum, S.P., et al., beta-catenin confers resistance to PI3K and AKT inhibitors and subverts FOXO3a to promote metastasis in colon cancer. Nat Med, 2012. 18(6): p. 892-901.

99

64. Lin, K., et al., daf-16: An HNF-3/forkhead family member that can function to double the life-span of . Science, 1997. 278(5341): p. 1319-22. 65. Tatar, M., et al., A mutant Drosophila homolog that extends life-span and impairs neuroendocrine function. Science, 2001. 292(5514): p. 107-10. 66. Holzenberger, M., et al., IGF-1 receptor regulates lifespan and resistance to oxidative stress in mice. Nature, 2003. 421(6919): p. 182-7. 67. Perez, V.I., et al., Protein stability and resistance to oxidative stress are determinants of longevity in the longest-living , the naked mole-. Proc Natl Acad Sci U S A, 2009. 106(9): p. 3059- 64. 68. Kops, G.J., et al., Forkhead transcription factor FOXO3a protects quiescent cells from oxidative stress. Nature, 2002. 419(6904): p. 316-21. 69. Nemoto, S. and T. Finkel, Redox regulation of forkhead proteins through a p66shc-dependent signaling pathway. Science, 2002. 295(5564): p. 2450-2. 70. Tran, H., et al., DNA repair pathway stimulated by the forkhead transcription factor FOXO3a through the Gadd45 protein. Science, 2002. 296(5567): p. 530-4. 71. Tothova, Z., et al., FoxOs are critical mediators of hematopoietic stem cell resistance to physiologic oxidative stress. Cell, 2007. 128(2): p. 325-39. 72. Paik, J.H., et al., FoxOs cooperatively regulate diverse pathways governing neural stem cell homeostasis. Cell Stem Cell, 2009. 5(5): p. 540-53. 73. Yalcin, S., et al., ROS-mediated amplification of AKT/mTOR signalling pathway leads to myeloproliferative syndrome in Foxo3(-/-) mice. EMBO J, 2010. 29(24): p. 4118-31. 74. Miyamoto, K., et al., Foxo3a is essential for maintenance of the hematopoietic stem cell pool. Cell Stem Cell, 2007. 1(1): p. 101-12. 75. Renault, V.M., et al., FoxO3 regulates neural stem cell homeostasis. Cell Stem Cell, 2009. 5(5): p. 527-39. 76. Warr, M.R., et al., FOXO3A directs a protective autophagy program in haematopoietic stem cells. Nature, 2013. 494(7437): p. 323-7. 77. Yeo, H., et al., FoxO3 coordinates metabolic pathways to maintain redox balance in neural stem cells. EMBO J, 2013. 32(19): p. 2589-602. 78. Rimmele, P., et al., Mitochondrial metabolism in hematopoietic stem cells requires functional FOXO3. EMBO Rep, 2015. 16(9): p. 1164-76. 79. Zhang, X., et al., FOXO1 is an essential regulator of pluripotency in human embryonic stem cells. Nat Cell Biol, 2011. 13(9): p. 1092-9. 80. Vilchez, D., et al., Increased proteasome activity in human embryonic stem cells is regulated by PSMD11. Nature, 2012. 489(7415): p. 304-8. 81. Goertz, M.J., et al., Foxo1 is required in mouse spermatogonial stem cells for their maintenance and the initiation of spermatogenesis. J Clin Invest, 2011. 121(9): p. 3456-66. 82. Ghaffari, S., et al., Cytokines and BCR-ABL mediate suppression of TRAIL-induced apoptosis through inhibition of forkhead FOXO3a transcription factor. Proc Natl Acad Sci U S A, 2003. 100(11): p. 6523-8. 83. Birkenkamp, K.U., et al., FOXO3a induces differentiation of Bcr-Abl-transformed cells through transcriptional down-regulation of Id1. J Biol Chem, 2007. 282(4): p. 2211-20. 84. Essafi, A., et al., Direct transcriptional regulation of Bim by FoxO3a mediates STI571-induced apoptosis in Bcr-Abl-expressing cells. Oncogene, 2005. 24(14): p. 2317-29. 85. Kharas, M.G., et al., Phosphoinositide 3-kinase signaling is essential for ABL oncogene-mediated transformation of B-lineage cells. Blood, 2004. 103(11): p. 4268-75. 86. Chapuis, N., et al., IkappaB kinase overcomes PI3K/Akt and ERK/MAPK to control FOXO3a activity in acute myeloid leukemia. Blood, 2010. 116(20): p. 4240-50.

100

87. Kornblau, S.M., et al., Highly phosphorylated FOXO3A is an adverse prognostic factor in acute myeloid leukemia. Clin Cancer Res, 2010. 16(6): p. 1865-74. 88. Scheijen, B., et al., FLT3 receptors with internal tandem duplications promote cell viability and proliferation by signaling through Foxo proteins. Oncogene, 2004. 23(19): p. 3338-49. 89. Sakoe, Y., et al., FOXO3A as a key molecule for all-trans retinoic acid-induced granulocytic differentiation and apoptosis in acute promyelocytic leukemia. Blood, 2010. 115(18): p. 3787-95. 90. Naka, K., et al., TGF-beta-FOXO signalling maintains leukaemia-initiating cells in chronic myeloid leukaemia. Nature, 2010. 463(7281): p. 676-80. 91. Sykes, S.M., et al., AKT/FOXO signaling enforces reversible differentiation blockade in myeloid leukemias. Cell, 2011. 146(5): p. 697-708. 92. Ross, M.E., et al., Gene expression profiling of pediatric acute myelogenous leukemia. Blood, 2004. 104(12): p. 3679-87. 93. Nagashima, T., et al., Discovery of novel forkhead box O1 inhibitors for treating : improvement of fasting glycemia in diabetic db/db mice. Mol Pharmacol, 2010. 78(5): p. 961-70. 94. Zhao, Y., et al., Cytosolic FoxO1 is essential for the induction of autophagy and tumour suppressor activity. Nat Cell Biol, 2010. 12(7): p. 665-75. 95. Mulloy, J.C., et al., The AML1-ETO fusion protein promotes the expansion of human hematopoietic stem cells. Blood, 2002. 99(1): p. 15-23. 96. Goyama, S., et al., UBASH3B/Sts-1-CBL axis regulates myeloid proliferation in human preleukemia induced by AML1-ETO. Leukemia, 2015. 97. Zhang, H., et al., Sox4 is a key oncogenic target in C/EBPalpha mutant acute myeloid leukemia. Cancer Cell, 2013. 24(5): p. 575-88. 98. Zeisig, B.B., et al., Hoxa9 and Meis1 are key targets for MLL-ENL-mediated cellular immortalization. Mol Cell Biol, 2004. 24(2): p. 617-28. 99. Essers, M.A., et al., Functional interaction between beta-catenin and FOXO in oxidative stress signaling. Science, 2005. 308(5725): p. 1181-4. 100. Elsasser, A., et al., The fusion protein AML1-ETO in acute myeloid leukemia with translocation t(8;21) induces c-jun protein expression via the proximal AP-1 site of the c-jun promoter in an indirect, JNK-dependent manner. Oncogene, 2003. 22(36): p. 5646-57. 101. Yamagata, K., et al., Arginine methylation of FOXO transcription factors inhibits their phosphorylation by Akt. Mol Cell, 2008. 32(2): p. 221-31. 102. Shia, W.J., et al., PRMT1 interacts with AML1-ETO to promote its transcriptional activation and progenitor cell proliferative potential. Blood, 2012. 119(21): p. 4953-62. 103. Mendler, J.H., et al., RUNX1 mutations are associated with poor outcome in younger and older patients with cytogenetically normal acute myeloid leukemia and with distinct gene and MicroRNA expression signatures. J Clin Oncol, 2012. 30(25): p. 3109-18. 104. Sattler, M., et al., Hematopoietic growth factors signal through the formation of reactive oxygen species. Blood, 1999. 93(9): p. 2928-35. 105. Hole, P.S., et al., Ras-induced reactive oxygen species promote growth factor-independent proliferation in human CD34+ hematopoietic progenitor cells. Blood, 2010. 115(6): p. 1238-46. 106. Eijkelenboom, A., et al., FOXO3 selectively amplifies enhancer activity to establish target gene regulation. Cell Rep, 2013. 5(6): p. 1664-78. 107. Monsalve, M. and Y. Olmos, The complex biology of FOXO. Curr Drug Targets, 2011. 12(9): p. 1322-50. 108. Martens, J.H., et al., ERG and FLI1 binding sites demarcate targets for aberrant epigenetic regulation by AML1-ETO in acute myeloid leukemia. Blood, 2012. 120(19): p. 4038-48. 109. Alic, N., et al., Interplay of dFOXO and two ETS-family transcription factors determines lifespan in . PLoS Genet, 2014. 10(9): p. e1004619.

101

110. Sartor, M.A., G.D. Leikauf, and M. Medvedovic, LRpath: A logistic regression approach for identifying enriched biological groups in gene expression data. Bioinformatics, 2009. 25: p. 211- 217. 111. Langmead, B. and S.L. Salzberg, Fast gapped-read alignment with Bowtie 2. Nat Methods, 2012. 9(4): p. 357-9. 112. Kumar, V., et al., Uniform, optimal signal processing of mapped deep-sequencing data. Nat Biotechnol, 2013. 31(7): p. 615-22. 113. Heinz, S., et al., Simple combinations of lineage-determining transcription factors prime cis- regulatory elements required for macrophage and B cell identities. Mol Cell, 2010. 38(4): p. 576- 89.

102

Chapter 4 Future implications

4.1 Future directions

Our study of MLL-AF4 leukemia demonstrates the complexity of modeling human disease, and it serves as a good example showing that the species differences between human and mouse have a significant role and cannot be ignored in acute leukemia studies. Follow-up studies are necessary to more fully understand the interesting questions raised in the current study. We have demonstrated that MLL-Af4 and

MLL-AF9 activate distinct target genes via differential DNA binding, however the molecular basis controlling the DNA binding preferences is not clear. A genome-wide ChIP-seq comparison may provide some insights. For example, if there are different recurrent binding motifs associating with MLL-Af4 and

MLL-AF9 binding sites, this could suggest that different co-factors recruit MLL-fusions and direct them to various targets. In addition, a comparative ChIP-seq study will tell us to what extent the genome-wide distributions of MLL-Af4 and MLL-AF9 are not overlapping, and how much of the transcription heterogeneity of MLL-r leukemia is contributed by differential DNA binding. Since both MLL-fusion proteins contain the same portion of MLL, the DNA binding specificity is likely controlled by the fusion partners. One hypothesis is that a fraction of the targets of wildtype MLL can be regulated by all MLL- fusion proteins and this contributes to the common signature of MLL-r leukemia. Meanwhile, MLL- fusion protein could bind and regulate the targets of the native fusion partners, leading to the diversity of

MLL-r disease. Most Af4 knockout mice can survive into adulthood but display impaired development of

B and T cells. Interestingly, the pre-B and mature B cell numbers are more severe reduced in Af4-/-mice compared to B progenitors at an earlier developmental stage [1]. By contrast, Af9-/-mice died soon after birth. It was reported the hematopoiesis appeared normal in newborn Af9-/- animals. Consistent with the finding that Af9 is highly expressed in the developing skeleton during embryogenesis, Af9 deficient mice exhibited homeotic transformations of the axial skeleton, and Hoxd4 was identified as an Af9 target[2].

Therefore, the mouse genetic studies indicate that Af4 and Af9 indeed regulate different targets and pathways, which may contribute to the phenotypic and transcriptional heterogeneity of the MLL-r

103 leukemia. A better understanding of binding sites and target genes of AF4 and AF9 in human normal hematopoiesis will provide invaluable clues.

Recruitment of different co-factors can led to different transcription outcomes even when the binding sites of two MLL-fusions are the same. Previous studies have suggested that MLL-fusion proteins could recruit various subtypes of SEC or DOT1L complex, and changing the composition of SEC could lead to non- identical functions[3-5]. However, most of these studies were not performed in leukemia cells, thus the complex conformation of the MLL-fusion protein in its native cellular environment is not clear. We have successfully immunoprecipitated the fusion proteins in ChIP assays, which will also allow us to investigate the interacting proteins of each fusion. We would expect to identify the specific co-factors associating with different fusions. In addition, as people have demonstrated that disruption of the protein- protein interaction can serve as a therapeutic strategy to fight against MLL-r leukemia, a detailed picture of the interactome of MLL-Af4 can provide novel therapeutic targets for this poor disease.

Immunotherapy has emerged as a powerful weapon against ALL. However we and others have found lineage switch was utilized by MLL-r cells to evade the therapy, reminding us there are no “magic bullets” that can eliminate all cancers. Fully understanding the molecular driving force of the lineage transition may help us to develop combined therapy to overcome the resistance. It would be interesting to test if our MLL-AF4 leukemia model will undergo similar lineage switch exposed to immunotherapy pressure. Considerations include that infusion of CAR-T cells into mice might induce Graft-versus-host disease (GVHD), but encouragingly, several recent studies showed the feasibility of infusing CAR-T cells into immunodeficient mice without GVHD induction and with efficacy against tumor cells in vivo[6, 7].

Therefore, it would be useful to establish a pre-clinical model of immunotherapy for our MLL-Af4 ALL cells, seeking the solution to improve the power of current therapy.

In the second study, we showed that a traditional tumor suppressor FOXO1 can act as an oncogene in t(8;21) leukemogenesis. This conflicting role has also been reported for other tumor suppressor genes, such as RUNX1 (AML1). RUNX1 has tumor repressor activity in several solid tumors[8, 9], and

104 inactivating RUNX1 mutations have been frequently found in a variety of myeloid disease[10]. However, we and others have shown that native RUNX1 is required for the survival of AML cells expressing MLL- fusion or AML1-ETO oncoproteins[11, 12]. Moreover, there are studies suggesting RUNX1 can serve as an oncogene in solid tumors[9, 13]. These data claim that there is no absolute separation between tumor suppressor and oncogene, with the function manifested by a gene being cancer subtype and cell context dependent. This concept should be emphasized when we develop and evaluate targeted therapy, as the heterogeneous nature of the leukemia could significantly skew the outcome. Although FOXO3 is required to maintain AML leukemia stem cells, it did not display oncogenic properties in our experiments. This result suggests that the leukemia stem cell programing and maintenance are two different functional aspects. The major role of FOXO3 in LSC is likely to be maintenance, and FOXO3 might coordinate the

ROS homeostasis, mitochondria function and metabolism of LSC in the same way as it works in normal

HSC. In contrast, FOXO1 is required for programing the cells and transcriptionally activating the self- renewal network. Accordingly, studies in both human ESC and mouse spermatogonial stem cells showed that those critical self-renewal genes were targets of FOXO1[14, 15]. Thus activating a stem cell program via FOXO1 might be a conserved mechanism used by a variety of stem cells. Future research analyzing and comparing the targets of FOXO1 and FOXO3 in both normal HSC and leukemia stem cells will help us to understand how each FOXO protein regulates the biology of leukemia stem cells.

The FOXO1 study also demonstrates the capability of the human model system. The typical molecular and biochemical assays can be easily performed on cells generated using this system, and the phenotype can be evaluated both in vivo and in vitro. There is no particular technical difficulties for researchers transiting from syngeneic mouse models to the human model, and the recent introduction of genome editing technology will open more possibilities using the human model system to investigate molecular and genetic interactions. For instance, we have shown that AE binds to a region within intron 1 of the

FOXO1 locus. We can therefore use CRISPR/Cas9 to delete this region in order to test if this region is a potential enhancer critical for FOXO1 activation in AE cells. One limitation of the current study is the

105 fact that thus far t(8;21) AML patient samples cannot be well supported in either in vitro culture systems nor in immunodeficient mice, and we therefore do not have an experiment system available to validate if

FOXO1 is required for pre-leukemia and leukemia cells in patients. We and others have shown that AE cells rely on cytokine THPO for survival, which does not have cross-species reactivity[16, 17]. The recent new immunodeficient mouse strain MISTRG has knock-in of human THPO gene[18]. Thus it is worth examining whether MISTG mice can better host t(8;21) patient samples to establish a pre-clinical model of AE leukemia, by which we can test whether targeting FOXO1 could facilitate the eradication of AE pre-leukemia and leukemia stem cells that survive chemotherapy.

Prognosis of certain subtypes of AML and ALL remain poor, and insight into the molecular mechanisms of the disease and the discovery of new therapeutic targets will help to improve the outcome. Disease models derived from primary human cells together with advances in humanized mouse models will continue to contribute significantly to the understanding of human leukemogenesis, leading to promising clinical applications.

4.2 References

1. Isnard, P., et al., Altered lymphoid development in mice deficient for the mAF4 proto-oncogene. Blood, 2000. 96(2): p. 705-10. 2. Collins, E.C., et al., Mouse Af9 is a controller of embryo patterning, like Mll, whose human homologue fuses with Af9 after chromosomal translocation in leukemia. Mol Cell Biol, 2002. 22(20): p. 7313-24. 3. Biswas, D., et al., Function of leukemogenic mixed lineage leukemia 1 (MLL) fusion proteins through distinct partner protein complexes. Proc Natl Acad Sci U S A, 2011. 108(38): p. 15751-6. 4. He, N., et al., Human Polymerase-Associated Factor complex (PAFc) connects the Super Elongation Complex (SEC) to RNA polymerase II on chromatin. Proc Natl Acad Sci U S A, 2011. 108(36): p. E636-45. 5. Lu, H., et al., Gene target specificity of the Super Elongation Complex (SEC) family: how HIV-1 Tat employs selected SEC members to activate viral transcription. Nucleic Acids Res, 2015. 43(12): p. 5868-79. 6. Huang, X., et al., IGF1R- and ROR1-Specific CAR T Cells as a Potential Therapy for High Risk Sarcomas. PLoS One, 2015. 10(7): p. e0133152. 7. Shiina, S., et al., CAR T Cells Targeting Podoplanin Reduce Orthotopic Glioblastomas in Mouse . Cancer Immunol Res, 2016. 4(3): p. 259-68. 8. Chimge, N.O. and B. Frenkel, The RUNX family in breast cancer: relationships with estrogen signaling. Oncogene, 2013. 32(17): p. 2121-30.

106

9. Ito, Y., S.C. Bae, and L.S. Chuang, The RUNX family: developmental regulators in cancer. Nat Rev Cancer, 2015. 15(2): p. 81-95. 10. Goyama, S., et al., Posttranslational modifications of RUNX1 as potential anticancer targets. Oncogene, 2015. 34(27): p. 3483-92. 11. Ben-Ami, O., et al., Addiction of t(8;21) and inv(16) acute myeloid leukemia to native RUNX1. Cell Rep, 2013. 4(6): p. 1131-43. 12. Goyama, S., et al., Transcription factor RUNX1 promotes survival of acute myeloid leukemia cells. J Clin Invest, 2013. 123(9): p. 3876-88. 13. Li, H., et al., Runx1 contributes to neurofibromatosis type 1 neurofibroma formation. Oncogene, 2016. 35(11): p. 1468-74. 14. Zhang, X., et al., FOXO1 is an essential regulator of pluripotency in human embryonic stem cells. Nat Cell Biol, 2011. 13(9): p. 1092-9. 15. Goertz, M.J., et al., Foxo1 is required in mouse spermatogonial stem cells for their maintenance and the initiation of spermatogenesis. J Clin Invest, 2011. 121(9): p. 3456-66. 16. Pulikkan, J.A., et al., Thrombopoietin/MPL participates in initiating and maintaining RUNX1-ETO acute myeloid leukemia via PI3K/AKT signaling. Blood, 2012. 120(4): p. 868-79. 17. Chou, F.S., et al., The thrombopoietin/MPL/Bcl-xL pathway is essential for survival and self- renewal in human preleukemia induced by AML1-ETO. Blood, 2012. 120(4): p. 709-19. 18. Rongvaux, A., et al., Development and function of human innate immune cells in a humanized mouse model. Nat Biotechnol, 2014. 32(4): p. 364-72.

107