• No results found

Biological implications of distinctive H3K79me2 patterns in acute myeloid leukemia

N/A
N/A
Protected

Academic year: 2021

Share "Biological implications of distinctive H3K79me2 patterns in acute myeloid leukemia"

Copied!
150
0
0

Loading.... (view fulltext now)

Full text

(1)

BIOLOGICAL IMPLICATIONS OF DISTINCTIVE H3K79ME2 PATTERNS IN ACUTE MYELOID LEUKEMIA

by

MOLLY CHRISTINE KINGSLEY B.S. Nazareth College, 2013

A thesis submitted to the Graduate School of the

University of Colorado in partial fulfilment of the requirements for the degree of

Doctor of Philosophy Molecular Biology Program

(2)

ii This thesis

for the Doctor of Philosophy degree of Molly Christine Kingsley

has been approved for the Molecular Biology Program

by

Tatiana Kutateladze, Chair Craig Jordan

Patricia Ernst James Hagman Nichole Reisdorph

Kathrin Bernt, Advisor

(3)
(4)

iv

Kingsley, Molly Christine (PhD, Molecular Biology Program)

Biological Implications of Distinctive H3K79me2 Patterns in Acute Myeloid Leukemia Thesis directed by Assistant Professor Kathrin M. Bernt

ABSTRACT

Mutations in isocitrate dehydrogenase 1 and 2 (IDH1/2) are found in 20-30% of adult acute myeloid leukemia (AML). Wild type IDH1/2 produce -ketoglutarate (KG), an important co-factor for many enzymes including DNA, RNA, and histone demethylases. Mutant IDH1/2 gain the ability to reduce KG and produce the structurally similar 2-hydroxyglutarate (2HG), a competitive inhibitor of KG. A previously reported consistent epigenetic alternation observed upon introduction of mutant IDH1/2 or treatment with 2HG is an increase in histone 3 lysine 79 di-methylation (H3K79me2).

MLL-rearrangements (ML-r), are found in about 10-20% of adult AML. MLL-r encompass MLL-fusions (MLL-F) and MLL-partial tandem duplications (MLL-PTD), both of which share a dependency on the H3K79 methyltransferase DOT1L for H3K79me2 on target genes such as HOXA9 and MEIS1. Pharmacologic inhibition of DOT1L results in downregulation of MLL fusion/PTD target genes. Interestingly, 30% of MLL-PTD AML also contain a mutation in IDH1/2. Given this high rate of co-occurrence we hypothesized that mutant IDH1/2 would cooperate with MLL-r through an H3K79me2 mediated mechanism. Inhibitors for both DOT1L and mutant IDH1/2 are currently in clinical trials; therefore, it is important to understand how different oncogenes cooperate and which treatments are the most effective.

Using murine models and viral overexpression we combined mutant IDH1/2 and MLL-fusions or PTDs and found no evidence of cooperation in our models. In fact, when

(5)

v

mutant IDH1 was introduced into a fully established MLL-r AML, genes regulated by DOT1L were downregulated resulting in a significant growth impairment. Due to the fact that numerous enzymes are inhibited by 2HG, we overexpressed DOT1L in order to selectively increase H3K79me2 levels. Once again, a significant impairment of growth was observed. H3K79me2 ChIP-Seq revealed that DOT1L overexpression increased H3K79me2 peaks on non-target genes, blunting the difference between fusion targets and other genomic loci. Given the unexpected necessity of different H3K79me2 levels on MLL-F target and non-target genes, we analyzed global levels in a panel of patient samples. We found that patients with MLL-r had globally low H3K79me2 compared to other types of AML.

In summary, we find an unexpected antagonism between MLL-r and mutations in IDH1/2 that may at least in part be explained by the need for MLL-r to maintain globally low, but locus specific high levels of H3K79me2. AML is a diverse disease with many patients presenting with unique combinations of mutations. The work presented here highlights that just because two mutations are found together does not mean they will cooperate. Knowing how different mutations interact is important for the rational design of new therapies.

The form and content of this abstract are approved. I recommend its publication. Approved: Kathrin M. Bernt

(6)

vi

DEDICATION

I dedicate this thesis to my grandpa Bob. At 89 years old he couldn’t wait for me to publish my work so I could bring the article down to Florida and read it with him. He was interested in the work I was doing and wanted to learn more. And while he will never see my published work I know if he were still here, he would want to read this entire thesis too. He was a life-long learner, his library filled with books and educational movies on all topics ranging from religion to modern science to world history. He had a wooden plaque in his living room, it read “With intelligence and hard work you will go 90% of the way, but it takes perseverance to go 100% of the way” He lived by these words and I aspire to as well. I can only hope that when I’m 89 I will be as well read and still eager to learn as he was.

(7)

vii

ACKNOWLEDGEMENTS

The work presented here would not have been possible without the help of many, many people. I have been so lucky to not have to walk this long road to my Ph.D. alone. I have an amazing support system and I will be forever grateful.

Clara, you have supported me and been by my side during this entire incredible journey. You have been there when I needed to cry, when I needed to laugh, and when I needed to talk about the science. You have an amazing ability that allows you to de-convolute my ideas and make them clear and without you this thesis wouldn’t be half as well written as it is now. Thank you for your endless proofreading and edits. Thank you for being my rock. Thank you for always encouraging me and reminding me that I could succeed.

Mom and Dad, what do I even say? Words will never be able to convey my gratitude for your endless support during these past 6 years. It hasn’t always been easy, there were a lot of frustrated phone calls but you two always knew how to put life into perspective and make me feel better. You were always there to celebrate the successes too. Calling you and hearing your excitement when I told you about an experiment went well was sometimes even better than the experiment working. I know you didn’t always know what I was talking about, but you celebrated just the same. Thank you for supporting and believing in me, especially on the days when I didn’t believe in myself.

Jeremy, you were always there exactly when I needed you the most and you somehow always knew exactly what to say. You made the long road toward this Ph.D. better and for that I thank you. And even though you beat me and became Dr. Kingsley first you never once doubted that I would get there too.

(8)

viii

Kathleen, thank you for always being there for me. You have been such an incredible support during this time, I could never thank you enough. You are a constant source of inspiration and happiness.

Todd, Jenny, Christine, and Brandon, you are my second family, my home away from home. Thank you for making Colorado feel like home. Thank you also for allowing me to show up on your doorstep with virtually no warning when I just needed a weekend away from it all. Thank you for you unwavering support and encouragement.

Jenn, Patrick, Tanya, and Anthony. Without you all I would have never made it through graduate school. We were all in together and together we have all made. You all made grad school fun, you helped keep me grounded, and reminded me that there is life outside of the lab. Thank you, thank you all for helping me through this journey.

To the entire GradWriteSlack community, thank you all for the endless amount of support and encouragement. This thesis would never have been finished if it weren’t for the writing buddies I found in this community. Thank you all so much for lifting me up when writing was the hardest and being there to celebrate the success.

Kathrin, you have pushed me and challenged me. I have grown so much as a scientist since joining your lab and learned more about mentoring than I can ever fully thank you for. Thank you for taking a chance on me when I was first in grad school and thank you for continuing to believe in me.

Finally, thank you to everyone who has been a part of this journey. To my committee thank you all for your ideas and for helping me to grow so much as a scientist. Sabrena and Michele, thank you both so much for all of your help over the years. I hope you both know that the molecular biology program would not run without either of you.

(9)

ix

Table of Contents

CHAPTER ABSTRACT………...iii DEDICATION………..…v ACKNOWLEDGEMENTS……….vi LIST OF FIGURES………..xiii LIST OF TABLES………..xv LIST OF ABBREVIATIONS……….xvi I. INTRODUCTION ...1

Acute myeloid leukemia...1

DOT1L and H3K79 methylation ...2

The discovery of Dot1 ...2

Unique features of DOT1L and H3K79me2 ...4

Roles of Dot1 and H3K79 methylation ...6

MLL and MLL-Rearrangements ...8

Wild Type MLL1 ...8

Discovery of MLL-rearrangements in leukemia...9

HOX Genes...10

HOX genes in development ...10

HOX genes in leukemia ...12

(10)

x

Hox cluster expression in other AML subtypes ...15

Isocitrate dehydrogenase...17

IDH1/2 mutations in glioblastomas ...17

Mutant IDH1 produces 2HG ...20

IDH1/2 mutations in AML ...20

Biological effects of 2HG ...21

Leukemic co-drivers with mutant IDH1/2 ...22

II. MUTANT IDH1/2 AND MLL-REARRANGEMENTS DO NOT COOPERATE TO DRIVE LEUKEMOGENESIS……….25 Abstract ...25 Introduction ...25 Methods ...28 Virus production ...28 Cell culture ...28

Generation of murine leukemia ...29

Colony formation assay ...30

Cell growth and drug assays ...31

Competition assay ...31

2HG treatment ...31

Histone and protein isolation and western blots ...31

(11)

xi

Library preparation and RNA-sequencing ...33

RNA-seq analysis ...33

Results ...33

Mutant IDH1 and 2HG inhibit growth of established MLL-r leukemia ...33

IDH1R132H and MLL-AF9 do not cooperate to drive leukemogenesis when co-expressed ...41

IDH2R140Q does not help MLL-AF9 drive leukemia when present first...44

Introduction of IDH1R132H into MLL-AF9 AML decreases expression of DOT1L dependent genes ...52

Discussion ...55

III. MLL-F AML REQUIRES PRECISE H3K79ME2 LEVELS ...58

Abstract ...58

Introduction ...58

Methods ...60

Viral production ...60

Murine leukemia cells ...60

Competition assay ...60

Histone extraction and western blots ...61

H3K79me2 chromatin immunoprecipitation ...61

Library preparation and DNA-sequencing ...62

(12)

xii

RNA extraction, library prep, RNA sequencing, and RNA analysis ...63

Results ...63

Increased levels of H3K79me2 are not tolerated in MLL-AF9 AML...63

Overexpression of DOT1L does not affect HOXA9/MEIS1 leukemia growth ...68

DOT1L overexpression increases H3K79me2 on non-target genes ...68

Discussion ...71

IV. CHARACTERIZATION OF LOCAL AND GLOBAL H3K79ME2 LEVELS IN LEUKEMIAS WITH HIGH HOX EXPRESSION ...76

Abstract ...76

Introduction ...76

Methods ...79

Human samples ...79

Thawing patient samples ...79

Cell growth and DOT1L inhibition of patient samples ...79

Histone and protein isolation and western blots ...80

H3K79me2 ChIP, ChIP library prep and sequencing, and ChIP-seq analysis ...80

RNA Extraction and cDNA production ...80

RNA library prep and sequencing and RNA-Seq analysis ...80

Results ...80

(13)

xiii

Patients with mutations in IDH1/2 have high H3K79me2 levels on HOXA and

HOXB clusters ...84

Mutant IDH1/2 Have High HOXA and HOXB Cluster Expression ...89

DOT1L dependent leukemias have a wide range of global H3K79me2 levels ...92

Discussion ...92

V. DISCUSSION AND FUTURE DIRECTIONS ...96

REFERENCES ... 105

(14)

xiv

LIST OF FIGURES

Figure 1.1: Mixed lineage leukemia protein domains ... 11

Figure 1.2: Roles of wild type IDH1/2/3 ... 18

Figure 1.3: Enzymes inhibited by 2HG ... 23

Figure 2.1: IDH1R132C slows the growth of human the MLL-PTD cell line EOL-1. ... 35

Figure 2.2: IDH1R132C slows the growth of murine MLL-AF9 leukemia ... 36

Figure 2.3: 2HG specifically slows growth of MLL-F driven cell lines ... 39

Figure 2.4: MLL-AF9 and IDH1R132H do not cooperate to drive leukemogenesis when co-expressed ... 42

Figure 2.5: MLL-AF9 and IDH1R132H driven AML is not dependent on IDH1R132H ... 45

Figure 2.6: Idh2R140Q does not cooperate with MLL-AF9 to drive leukemogenesis in vivo ... 48

50 Figure 2.7: Idh2R140Q does not cooperate with MLL-AF9 to drive leukemogenesis in vitro ... 50

Figure 2.8: DOT1L program is down regulated when IDH1R132H is overexpressed in MLL-AF9 leukemia ... 53

Figure 2.9: H3K79me2 levels are unchanged in MLL-AF9 model or bone marrow ... 54

Figure 3.1: Experimental set up ... 64

Figure 3.2: Neither increased nor decreased levels of H3K79me2 are sustained in MLL-AF9 cells ... 66

Figure 3.3 Overexpression of DOT1L is not tolerated in MLL-AF9 leukemia cells. ... 67

Figure 3.4 Overexpression of high levels of DOT1L dead is not tolerated in MLL-AF9 leukemia cells. ... 69

(15)

xv

Figure 3.6: Overexpression of DOT1L results in increased H3K79me2 peaks outside of

typical genomic loci resulting in decreased expression. ... 72

Figure 4.1: Mutant IDH1/2 driven AML are dependent on DOT1L ... 82

Figure 4.2: H3K79me2 patterns in AML samples. ... 85

Figure 4.3: Specific patterns of H3K79me2 in mutant IDH1/2 patient samples. ... 87

Figure 4.5: Mutant IDH1/2 have high HOXA9 and HOXB4 expression... 90

(16)

xvi

LIST OF TABLES

Table A.1: Patient Characteristics………122 Table A.2: Reagents………..………130 Table A.3: Primers……….………132

(17)

xvii

LIST OF ABBREVIATIONS

2HG 2-hydroxyglutarate 5mC 5-methylcytosine 5hmC 5-hydroxymethylcytosine KG -ketoglutarate

AML Acute myeloid leukemia

DOT1(L) Disruptor of telomeric silencing 1 (Like)

DNMT3A DNA Methyltransferase 3A

IDH1 Isocitrate dehydrogenase 1

IDH2 Isocitrate dehydrogenase 2

ITD Internal tandem duplication

me1, me2 or me3 mono, di, or trimethylation

MLL Mixed lineage leukemia - 1

MLL-r Mixed lineage leukemia - 1 rearrangement

MLL-F Mixed lineage leukemia - 1 fusion

NPM1 Nucleophosmin

NPM1c Cytoplasmic Nucleophosmin

PTD Partial tandem duplication

SET Su(var) 3-9, Enhancer-of-zeste/Trithorax

(18)

1

CHAPTER I

INTRODUCTION

Over the past 50 years epigenetic modifications have emerged as critical regulators of biological functions. More recently, many studies have focused on the role of epigenetic modifiers in cancer. This thesis focuses on two epigenetic modifiers, mixed lineage leukemia (MLL) and isocitrate dehydrogenase (IDH) in acute myeloid leukemia (AML). A unifying feature of both is a dependency on the H3K79 methyltransferase, disruptor of telomeric silencing 1 like (DOT1L).

Acute myeloid leukemia

Long-term hematopoietic stem cells (LT-HSC), sit at the top of a hierarchy of increasing differentiated cells that ultimately make up the blood system. LT-HSC are located in the bone morrow and along with short-term hematopoietic stem cells (ST-HSC) and multi-potent progenitors (MPP) have both self-renewal and differentiation abilities. (Rieger and Schroeder, 2012). In normal blood development, cells systematically move down the hematopoietic tree becoming more lineage restricted at each stage. The different stages of development can be separated based on differing expression of cell surface markers (Sykes and Scadden, 2013).

Acute myeloid leukemia (AML) is defined as a clonal expansion of immature cells. The rapid uncontrolled proliferation of the immature cells, known as leukemia blasts, results in the loss of functional blood cells and ultimately a collapse of the myeloid lineage. There are a vast number of different mutations found in AML. Most, however, can be defined by contributing to unrestrained proliferation or inhibiting differentiation. (Donnell et al., 2011;

(19)

2

Estey and Do hner, 2006; Gilliliand and Griffin, 2002; Hope et al., 2004; Lowenberg et al., 1999; Shipley and Butera, 2009).

DOT1L and H3K79 methylation The discovery of Dot1

DNA is wound around an octamer of proteins, two of each of the histones; H2A, H2B, H3 and H4. DNA loops around each histone core 1.7 times, creating the nucleosome (Luger et al., 1997). Nucleosomes can be packed closely together or spread far apart; the proximity of nucleosomes creates a general organizational pattern within the nucleus, known as heterochromatin and euchromatin. Heterochromatic regions consist of nucleosomes that are packed tightly together, creating inaccessible DNA and silencing the genes. Nucleosomes in the euchromatic regions are widely spaced and the DNA is accessible to transcriptional machinery (Allshire and Madhani, 2018; Babu and Verma, 1987). The telomeres are a notable area of the DNA that is packed into heterochromatin. Telomeres, found at the end of linear DNA sequences consist of silent repeating sequences that slowly shrink with each cellular division and replication of the DNA.

Disruptor of telomeric silencing 1 (Dot1) was first discovered in 1998 in S. cerevisiae in a cDNA overexpression screen for proteins that disrupted heterochromatic DNA and telomeric silencing. In yeast, 10% of the DNA is tightly packed into heterochromatin; however, there are many genes that exist in open chromatin that are not actively transcribed. Overexpression and deletion of Dot1 severely impaired silencing of heterochromatic regions. From these results the authors speculated that Dot1 was a chromatin protein involved with repressive chromatin (Singer et al., 1998). After the initial discovery of Dot1, four groups independently found that Dot1 and Dot1L (Disruptor of Telomeric Silencing 1

(20)

3

Like, the mammalian homolog) methylated histone 3 at lysine 79 (H3K79) (Feng et al., 2002; Lacoste et al., 2002; Ng et al., 2002a; van Leeuwen et al., 2002). As Dot1 and Dot1L are highly homologous and function similarly, Dot1 will be used when referring to features common to both. Dot1L will be used when specifically referring to the mammalian protein.

One mechanism of heterochromatin silencing is facilitated by the silent information regulator (Sir) family of histone deacetylases. Acetylation neutralizes the positive charge of lysines resulting in a loosening of the DNA from histones and a decompaction of the chromatin (Heintzman et al., 2007; Schubeler, 2007; Schu beler et al., 2004). Thus, histone lysine acetylation has an important role in active transcription by keeping genes “open” for transcription machinery. The removal of the acetyl group then allows the DNA to tightly condense back around the histone and be silenced (Moazed, 2001).

Dot1 and H3K79 methylation were first proposed to encourage Sir binding and gene silencing. Loss of Dot1 or mutation of histone 3 K79 to A79 (which cannot be methylated) decreased binding of Sir2/3 at increasing distances away from the telomeres. These data suggested that H3K79methylation was important for Sir2/3 binding (Ng et al., 2002a). In important counterpoint to this hypothesis was the finding H3K79me2 covers 90% of the yeast genome, while Sir proteins only bind to 10%. Therefore, H3K79me2 could encourage Sir binding but is not sufficient for the recruitment (van Leeuwen et al., 2002).

In a further analysis of the function of DOT1L, it was found that both overexpression and deletion of Dot1 decreased telomeric silencing. In contrast what was first published, that overexpression and not loss of Dot1, decreased Sir2/3 binding. This suggests that Sir proteins bind preferentially to unmethylated K79. To reconcile the seemingly contradictory results that both overexpression and deletion of Dot1 could disrupt telomere silencing, the

(21)

4

following model was proposed. Overexpression of Dot1 increases H3K79me3 in areas where it is not normally found, such as in telomeric regions, therefore preventing Sir2/3 binding. On the other hand, by deleting Dot1 and decreasing H3K79 methylation, Sir2/3 binding promiscuously increases. The extended range of Sir2/3 binding, while resulting generally in more genomic silencing, also takes Sir2/3 away from the regular binding sites, thus disrupting telomeric silencing (van Leeuwen et al., 2002).

In further support of Sir preferentially binding to unmethylated histones, it was proposed that Dot1 and Sir2 function by way of mutually exclusive positive feedback loops (Ng et al., 2003). Overexpression of Dot1 increases H3K79me2 which inhibits Sir2 binding and allows for DOT1L to methylate more genes. On the other hand, increasing Sir2 binding precludes H3K79me2 and promotes more Sir binding in areas where there is no H3K79me2 (Ng et al., 2003). This paper added the final evidence to the field that Dot1 and Sir2 operate in a mutually exclusive manner thereby balancing transcription and silencing. Despite the discovery of Dot1 as a disruptor of silencing and its relationship to silencing, all of the aforementioned papers came to the conclusion that Dot1 and H3K79me2/3 are associated with active transcription.

Unique features of DOT1L and H3K79me2

Histone lysine methylation modifications are found predominately on the tails of histone 3. Lysine residues can be mono, di or tri methylated (me1, me2, me3). Unlike acetylation, which is always found on active genes, lysine methylation is associated with both active and silent genes. While methylation on histone 3 lysines 9 and 27 and histone 4 lysine 20 (H3K9, H3K27, H4K20) are associated with silent genes, methylation on histone 3 lysines

(22)

5

4, 36 and 79 (H3K4, H3K36, K3K79) are associated with active genes (Bannister and Kouzarides, 2011).

Numerous features make Dot1 unique from the other previously characterized methyltransferases. First, Dot1 methylates lysine 79 which is found on the globular surface of histone 3, not on the tail where lysines 4, 9, 27, and 36 are all located (Feng et al., 2002; Lacoste et al., 2002; Ng et al., 2002a; van Leeuwen et al., 2002). Due to this different location, it follows that Dot1 utilizes a different type of methyltransferase domain. Lysine methylation is a stepwise process achieved though the binding of a S-adenosyl methionine (SAM) molecule, which provides a methyl group (Black et al., 2012). All of the other histone methyltransferases discovered to date, use a Su(var) 3-9, Enhancer-of-zeste/Trithorax (SET) domain for methylation. DOT1L instead has a methyltransferase fold, a domain that is similar to those found in arginine methyltransferases (Ali et al., 2013; Black et al., 2012; Cuthbert et al., 2004; Feng et al., 2002; Min et al., 2003; Ng et al., 2009; Nguyen and Zhang, 2011).

Another specificity of Dot1 is that it is unable to methylate free histones, instead an intact nucleosome must be provided. It was found that adding increasing amount of DNAse to a solution of nucleosomes and Dot1 gradually decreased the activity of Dot1. While a single histone or even a histone tail is a sufficient substrate for many of the other methyltransferases, by needing an intact nucleosome this result suggests that Dot1 requires inter-histone communication to function (Feng et al., 2002; Lacoste et al., 2002; Min et al., 2003; Ng et al., 2002a; Park et al., 2002; van Leeuwen et al., 2002). Concordantly, ubiquitination of histone H2B at lysine 123 (H2BK123ub) is a prerequisite for di and tri methylation of H3K79 by Dot1 (Anderson et al., 2019; Briggs et al., 2002; Nakanishi et al., 2009; Ng et al., 2002b; Schulze et al., 2009; Wysocki et al., 2005; Yao et al., 2019).

(23)

6

The final major difference is that a demethylase for H3K79 has not yet been discovered. There are two families of lysine demethylases; lysine specific demethylases (LSD) and Jumonji containing demethylases (JmjC). The LSD family can only demethylate mono and di methylated histones (Hou and Yu, 2010; Shi et al., 2004). In contrast, JmjC demethylases can remove all levels of lysine methylation. Demethylation by the JmjC family is achieved through an oxidative decarboxylation reaction. JmjC demethylases are part of a large family of oxygenases that require iron and -ketoglutarate (KG) (Chen et al., 2006; Klose et al., 2006; Shi and Tsukada, 2013; Tsukada et al., 2006; Whetstine et al., 2006).

While the existence of an H3K79 demethylase remains a controversy in the field, there are several lines of evidence supporting its existence. First, the levels of H3K79me2 have been shown to be closely linked to cell cycle, with H3K79me2 levels peaking during M phase and rapidly decrease during G2, a pattern that cannot be explained by only nucleosome turnover (Ooga et al., 2008; San-Segundo and Roeder, 2000; Schulze et al., 2009). Second, in a study of H3K79me2 levels in oocyte fertilization, all H3K79me2/3 is rapidly lost after an oocyte has been fertilized in flies and mice. The vast majority of other histone methylation marks are also lost during this time through active demethylation, thus suggesting H3K79me2/3 would also be removed by a demethylase (Ooga et al., 2008; Shanower et al., 2005). Finally, when 2HG, a known inhibitor of JmjC demethylases was added to gliomas cells a rapid increase of H3K79me2 was observed, suggesting a demethylase was inhibited (Xu et al., 2011a).

Roles of Dot1 and H3K79 methylation

Dot1 is involved with numerous functions in the cell including cell cycle regulation (Mellor, 2009; San-Segundo and Roeder, 2000; Wysocki et al., 2005), DNA damage response

(24)

7

(Wysocki et al., 2005), embryonic development (Jones et al., 2008; Ooga et al., 2008; Shanower et al., 2005), cardiomyocyte differentiation (Cattaneo et al., 2014), erythropoiesis (Feng et al., 2002), transcriptional regulation (Bitoun et al., 2007; Mueller et al., 2007), and leukemogenesis (Bernt et al., 2011). This thesis will be focusing on the roles Dot1 has in transcriptional regulation and how the roles relates to the function of DOT1L in leukemogenesis.

While about 10% of the mammalian genome is H3K79 methylated at any given time, complete loss of DOT1L only results in transcriptional changes in a small number of genes (Bernt et al., 2011; Jones et al., 2008; van Leeuwen et al., 2002). This suggests that H3K79me2 is not essential for most of transcription. Similar to what was discovered in yeast, H3K79me2 helps selectively maintain a state of open chromatin in leukemia by inhibiting the function of the SIRT1, mammalian homolog of the yeast histone deacetylase Sir2 (Chen et al., 2015).

Which complexes DOT1L is a part of has long been a subject of debate. One of the first complexes DOT1L was found in is the ENL-associated protein (EAP) complex, which contains ENL, DOT1L, AF4, AF5q31, LAF4, Ring1, CBX8, BCoR, and P-TEFb (Mueller et al., 2007). A second, similar complex, containing DOT1L, AF4, AF9, ENL, AF10, and P-TEFb was identified soon after (Bitoun et al., 2007). It is thought that DOT1L is brought into the complexes through direct interactions with ENL or AF10 (Okada et al., 2005; Okada et al., 2006; Zeisig et al., 2005). A follow-up study by Mueller et al. found that not all of the originally predicted proteins could be together all together all the time. They therefore revised the complex to the EAP core which only consisted of DOT1L, ENL, P-TEFb, and AF4 (Mueller et al., 2009). Nonetheless, one of the key proteins found in both of the above complexes is P-TEFb, a cyclin

(25)

8

dependent kinase that is responsible for phosphorylating the C-terminal domain of RNA-polymerase II. The phosphorylation event, releases RNA-RNA-polymerase II from its stalled state and allows transcription to begin. In another study, on the interaction partners of DOT1L it was found that the core DOT1L complex (DOTcom) consists of DOT1L, AF10, AF17, and AF9/ENL, but not P-TEFb (Mohan et al., 2010). In all of the above-mentioned complexes many of the interacting partners are transcription factors, and interestingly have been found to be involved in fusion with the mixed lineage leukemia (MLL) protein.

MLL and MLL-Rearrangements Wild Type MLL1

Mixed lineage leukemia (MLL1), also known as homolog of trithorax (HRX) or lysine methyltransferase 2A (KMT2A), is a member of a family of H3K4 methyltransferases. H3K4me1/2/3 are all strongly associated with active transcription, with H3K4me1 found primarily at enhancers and H3K4me3 found at gene promoters. There are six proteins in the family, MLL1, MLL2, MLL3, MLL4, SET1A, SET1B (Crump and Milne, 2019). This thesis will be focusing on MLL1, which is found at chromosomal band 11q23 and is involved with leukemogenic rearrangements. In addition to the H3K4 methyltransferase domain, MLL1 also contains three AT hooks and a CxxC domain for DNA binding, four PHD domains for protein-protein interactions, a bromodomain that binds to acetylated histones, and a transactivation domain that interacts with the transcriptional co-activator, CBP (Birke et al., 2002; Ernst et al., 2001; Zeleznik-Le et al., 1994). MLL interacts with RbBP5, Ash2L and WDR5 to form the core of the COMPASS-like complex (Dou et al., 2006). Additional interaction partners include Menin and MOF (Caslini et al., 2007; Yokoyama and Cleary,

(26)

9

2008). MLL1 is a master transcriptional regulator that controls expression of the HOX genes during development (Milne et al., 2002).

Discovery of MLL-rearrangements in leukemia

Throughout the 1980s numerous papers were published describing a rearrangement found at chromosomal band 11q23 (Daeschner et al., 1985; Kaneko et al., 1986; Raimondi et al., 1989a; Raimondi et al., 1989b). Due to the reoccurring nature of the 11q23 rearrangement in leukemia it was hypothesized that the gene at 11q23 was involved with leukemogenesis. In 1992, the gene found at 11q23 was cloned and found to have high sequence homology with the Drosophila gene trithorax, thus it was named homolog of trithorax (HRX). Trithorax is a master transcription regulator responsible for the expression of the HOX genes, genes involved in development and body patterning in Drosophila. Sequence homology was particularly high at the C-terminal domain, as well as at the zinc-finger motifs. Additionally the gene involved in the t(11;19) rearrangement was cloned and named eleven-nineteen leukemia (ENL) (Tkachuk et al., 1992). In 1993, a second group also cloned the gene at 11q23 and they named it MLL for mixed lineage leukemia because patients with it often displayed characteristics of both myeloid and lymphoid leukemia. They mapped the breakpoint region to a 8.3kb area between exon 9 and 11 and documented 10 different fusion partners found in patients (Thirman et al., 1993).

Today, over 100 different fusion partners have been identified; however, six make up the vast majority of MLL-Fusions (MLL-F): AF4, AF6, AF9, AF10, ENL, and ELL. Most of the common wild type fusion partners are found in the transcriptional complexes AEP or SEC (Meyer et al., 2013). In the leukemogenic protein fusions, the C-terminal end of MLL encompassing the methyltransferase, PHD, and bromodomains is lost, leaving the

(27)

N-10

terminal end with the AT hooks and CxxC DNA binding domains. In leukemia, MLL-F are directed to many of the targets of WT MLL, including HOX genes, resulting in their overexpression.

Another type of rearrangement of MLL found in leukemia is the partial tandem duplication (MLL-PTD). In contrast to the MLL-F, in the MLL-PTD the entire protein is retained and exon 2 through 8-11 are duplicated (Basecke et al., 2006). Throughout this thesis, I am referring to MLL-F and MLL-PTD collectively as MLL-rearrangements (MLL-r) (Figure 1.1) One major difference between MLL-F and MLL-PTD is in the number of co-occurring mutations. While MLL-F are exceptionally strong oncogenes that are able to drive leukemia with little to no other mutations, MLL-PTD tend to occur with additional mutations (Andersson et al., 2015; Sun et al., 2016).

HOX Genes

HOX genes in development

Homeobox-containing genes (Hox) encode for a large family of transcription factors with an evolutionarily conserved DNA-binding domain. HOX proteins control a multitude of genes relating to cell cycle, other transcription factors, oncogenes and tumor suppressors. HOX proteins are capable of activating or silencing genes (Dorsam et al., 2004; Lappin et al., 2006; Lewis, 1978). The Hox family of genes were first identified in Drosophila as controllers of body patterning and development. Early studies of the Hox genes in Drosophila found that mutation of different genes severely impaired development. In a series of landmark studies mutation of the antennapedia gene resulted in legs growing in place of antennae (Bournias-vardiabasis and Bownes, 1978; Lewis, 1978; Postlehwait and Schneiderman, 1971). In

(28)

11

Drosophila, there is only one cluster of Hox genes, while in many mammals there are four HOX clusters, A-D, found on different chromosomes.

Figure 1.1: Mixed lineage leukemia protein domains

A) Wild type MLL contains AT hooks and CxxC domains for DNA binding, PHD and bromodomains (BD) involved in protein-protein interactions, and a SET domain that methylates H3K4.

B) In MLL-Fusions the PHD, BD, and SET domains are lost and replaced with 1 of over 100 different fusions partners.

(29)

12

Each cluster has 9-11 genes whose expression is coordinated with development and body orientation. The low number, 3’ Hox genes, are expressed in the anterior part of the body early on, while high number, 5’ genes, are expressed in the posterior part of the body later in development. This effect is referred to as temporal collinearity (Lappin et al., 2006).

Temporal collinearity is not restricted to general body patterning but is also found in the hematopoietic system. In mammals, there is stepwise expression of the Hoxa, Hoxb and Hoxc clusters in hemopoietic development. The Hoxa cluster is highly expressed in stem cells, with the 3’ half of the cluster expressed in the most primitive cells and the 5’ half expressing later in the hematopoietic lineage. Hoxb cluster genes are also expressed in stem cells, but remain on later in development when most of the Hoxa cluster genes have turned off (Giampaolo et al., 1994; Magli et al., 1991; Moretti et al., 1994; Sauvageau et al., 1994). In leukemia, oncogenes are able to hijack this system. Re-expression of Hox cluster genes that should be silenced will result in a block in differentiation.

HOX genes in leukemia

HOXA gene overexpression was first implicated in leukemogenesis by Neal Copeland’s group in 1996. Using proviral tagging in BXH-2 mice Hoxa9, Hoxa7 and murine ectopic integration site 1 (Meis1) were identified as potential oncogenes. In this system, viral integration is used to identify genes, that when expressed, contribute to leukemia in mice. In 95% of the mice, where Meis1 expression was activated by proviral integration, either Hoxa9 or Hoxa7 were also expressed. This strongly suggested a cooperative mechanism (Nakamura et al., 1996).

Over time many analyses of human leukemias have added evidence of the cooperative and transforming nature of MEIS1 and HOXA genes (Armstrong et al., 2002; Ayton and Cleary,

(30)

13

2003; Dorrance et al., 2006; Ferrando et al., 2003; Milne et al., 2005a; Okada et al., 2005; Okada et al., 2006; Zeisig et al., 2003). While, overexpression of HOXA9 and MEIS1 are found across many subtypes of leukemia, they are most famously known for their role in MLL-r leukemia (Rozovskaia et al., 2001).

Intertwined roles of DOT1L, MLL and HOX genes

Two features have come to define MLL-r leukemia; high expression of HOXA7- HOXA10 genes (Armstrong et al., 2002; Ayton and Cleary, 2003; Dorrance et al., 2006; Ferrando et al., 2003; Milne et al., 2005a; Okada et al., 2005; Okada et al., 2006; Zeisig et al., 2003) and a dependency on DOT1L and H3K79me2 (Bernt et al., 2011; Garcia-Cuellar et al., 2016; Kerry et al., 2017; Okada et al., 2005; Okada et al., 2006). In 2005, Yi Zhang’s group first linked DOT1L and MLL-F. Using a yeast-two-hybrid screen they found the C-terminal end of AF10 was an interaction partner of Dot1l. The interaction was mapped to the octapeptide motif (OM) and leucine zipper (LZ) domains of AF10 (Okada et al., 2005), the same domain that was previously found to be critical for transformation by MLL-AF10 (Chaplin et al., 1995; DiMartino et al., 2002). Okada et al showed that the Dot1l/AF10 interaction was required for transformation and that Dot1l mediated H3K79 methylation of Hoxa9 was a major driver of Hoxa9 overexpression (Okada et al., 2005). In their next paper, they showed that Dot1l was also required for CALM-AF10 driven AML. Dot1l binding to AF10 helped CALM-AF10 localize to the nucleus. CALM is a cytoplasmic protein and without DOT1L directed the fusion protein out of the nucleus. Additionally, DOT1L was responsible for the methylation of Hoxa5 which contributed to the high expression level found in CALM-AF10 AML (Okada et al., 2006).

(31)

14

Around the same time Jay Hess’ lab provided evidence of the importance of DOT1L in MLL-ENL AML. They created a model where expression of MLL-ENL was controlled by ER-CRE. When MLL-ENL was removed and then subsequently re-expressed, H3K79me2 was the most changed histone modification. Furthermore, H3K79me2 was strongly correlated to rapid expression changes in the Hoxa cluster and Meis1 (Milne et al., 2005b). H3K4me3 was modestly affected by MLL-ENL expression but had a low correlation with Hoxa cluster and Meis1 expression. Interestingly they also found strong overlap between where MLL-ENL was binding in the genome and H3K79me2. They hypothesized that binding of MLL-F would decrease binding of WT MLL; however, opposite was found. Binding of WT MLL slightly increased in the presence of MLL-ENL and together the two regulated HOX gene expression (Milne et al., 2005b). At the time, this led many to believe that MLL-fusions inappropriately recruited Dot1L to MLL targets such as the Hoxa cluster, resulting in inappropriate expression.

With evidence steadily mounting that DOT1L was important for MLL-r AML and HOX expression, using a Dot1l knockout mouse, it was found that while H3K79me2 was globally lost, it was predominantly the MLL-F targets that changed expression. The MLL-F target genes were then defined as those where MLL-AF9 was bound. Furthermore, Bernt et al showed that the H3K79me2 pattern was different on the target genes than other genes. On the MLL-F target genes H3K79me2 covered the entire gene body, while on other highly expressed genes H3K79me2 peaked after the transcription start site and decreased over the first exon. The dependence of MLL-r AML on DOT1L was further confirmed using a small molecule inhibitor of DOT1L (Bernt et al., 2011). Subsequent studies have also defined target genes for the various fusion partners. While some differences are found, HOXA7-HOXA10 and

(32)

15

MEIS1 are part of a common core of genes regulated by MLL-F in AML (Garcia-Cuellar et al., 2016; Kerry et al., 2017). Additional characteristics of MLL-F target genes are high levels of H3K79me2 and decreased expression upon loss of H3K79me2.

All of the initial studies linking DOT1L and MLL-r were performed using nuclear fusion partners; therefore, it was surprising when the fusion to a cytoplasmic protein (MLL-AF6) and MLL-PTD were also found to be dependent on DOT1L (Deshpande et al., 2013; Ku hn et al., 2015a; Ku hn et al., 2015b). A complete mechanism of why these leukemias are dependent on DOT1L is an area of active investigation.

Hox cluster expression in other AML subtypes

The expression of different HOX genes can be used to classify different subtypes of AML (Armstrong et al., 2002). Leukemic subtypes can be classified into four categories based upon whether they express genes from only the HOXA cluster, HOXB cluster, both, or neither (Spencer et al., 2015). A subtype of AML that expressed genes from both clusters were those with cytoplasmic nucleophosmin mutations (NPM1c) (Spencer et al., 2015). NPM1 is normally a nuclear protein; responsible for controlling transcription factors such as PU.1. In AML, mutations forces NPM1 to localize to the cytoplasm where it no longer is able to control transcription, resulting in a block in differentiation and uncontrolled proliferation (Gu et al., 2018; Verhaak et al., 2005). It was also notable that many of the proteins that affect epigenetic modifications in some way: DNA methyltransferase 3A (DNMT3A), ten-eleven-translocation (TET2), and isocitrate dehydrogenase 1 and 2 (IDH1/2) were also in the category with high expression of both HOXA and HOXB genes (Spencer et al., 2015).

DNMT3A and TET2 are both major regulators of DNA methylation. DNA methylation (5mC) occurs on the 5’ position of the cytosine ring (5mC) and is a key mechanism cells use

(33)

16

to silence genes and is a major factor in X-chromosome inactivation (Edwards et al., 2017). Promoter regions have especially high concentrations of DNA methylation compared to the rest of the genome.

The DNA methyltransferase family of proteins are responsible for placement of methylation. DNMT3A and DNMT3B both perform de novo cytosine methylation while DNMT1 is responsible for maintenance methylation after DNA replication. In AML, inactivating mutations in DNMT3A lead to global hypomethylation, gene activation, and increased proliferation (Brunetti et al., 2017; Challen et al., 2012; Cullen et al., 2014; Edwards et al., 2017). Loss of DNA methylation on the HOX cluster may be one mechanism by which the genes are upregulated in AML.

The TET family of proteins are responsible for the removal of methylation on DNA. DNA demethylation is an KG dependent multistep process (Bhutani et al., 2011). Originally it was thought that the intermediates were transient and therefore had no biological function; however, studies over the past decade have begun to provide evidence to the contrary. The most well studied of the intermediates is 5-hydroxymethylcytosine (5hmC). Oxidation of 5mC by TET2 directly produces 5hmC, which has been found on enhancers and distinctive active gene programs (Johnson et al., 2016; Pfeifer et al., 2013; Stroud et al., 2011; Wossidlo et al., 2011). In AML, inactivating mutations in TET2 lead to a global increase of 5mC and a subsequent decrease in 5hmC. The resulting hypermethylation silences many genes and increasing stem cell self-renewal ability (Akalin et al., 2012; Cimmino et al., 2017; Figueroa et al., 2010a; Figueroa et al., 2010b; Moran-Crusio et al., 2011).

An important commonality between NPM1c, DNMT3A mutants and MLL-r is they are all dependent on DOT1L and H3K79me2. This raises the question, are all leukemias with high

(34)

17

expression of HOX cluster genes dependent on DOT1L? As described in chapter IV we find this may be the case.

Isocitrate dehydrogenase

Wild type isocitrate dehydrogenase (IDH) is responsible for the conversion of isocitrate to -ketoglutarate (KG). There are three isoforms of IDH, IDH1 is found in the cytosol; IDH2 and IDH3 are found in the mitochondria. IDH3 is a part of the Krebs cycle and, to date, has not been found to be mutated in cancer. IDH1 and IDH2, which have both been found to be mutated in cancer, participate in various metabolic pathways (Figure 1.2). However, it is the product of IDH that is of interest for this thesis, KG is not only an intermediary building block for amino acids and fatty acids but is also is an essential co-factor for a broad class of enzymes that catalyze hydroxylation reactions. Included in this family are histone demethylases, nucleotide hydroxylases (DNA and RNA demethylases), and prolyl hydroxylases. Single point mutations in the active sites of IDH1 or IDH2 confer the ability to reduce KG to 2-hydroxyglutarate (2HG), a structurally similar molecule that is a competitive inhibitor for all enzymes that rely on KG. Thus, mutant IDH1 and IDH2 affect all layers of epigenetic modifications, and multiple metabolic and signaling pathways (Dang et al., 2009; Gross et al., 2010; Johansson et al., 2014)

IDH1/2 mutations in glioblastomas

Mutations in IDH1 were first found in a large screen of colorectal and breast cancers. A single patient with colorectal cancer had a mutation in IDH1 at arginine 132 (R132) (Sjoblom et al., 2006). Two years later in a large-scale genomic analysis of glioblastoma multiforme patients’ heterozygous mutations in IDH were found in 12% of the patients. All mutations were at R132 and the vast majority were mutated to histidine.

(35)

18 Figure 1.2: Roles of wild type IDH1/2/3

IDH2 and 3 are both located in the mitochondria and IDH1 is located in the cytosol. IDH3 functioning in the Krebs cycle and has not been implicated in any cancers. IDH2 and IDH1 both provide KG for amino acid and fatty acid synthesis.

IDH3 Citrate aKG Krebs Cycle Isocitrate Isocitrate aKG Glutamate Isocitrate aKG IDH2 IDH1 Citrate

AcetylCoA Fatty acids

Mitochondria

(36)

19

They noted that mutations in IDH1 were found in younger patients with secondary glioblastoma multiforme and these patients had better outcomes than patients with WT IDH1 (Parsons et al., 2008). Soon after, two studies were published demonstrating that mutations in IDH1 were found early in cancer progression, strongly suggesting mutant IDH1 is an initiating mutation in brain tumors. Early and late biopsies revealed that mutations in IDH1 were often detected in the first biopsy and always appeared before TP53 mutations. Both studies also extended the types of patients included in the analysis and found IDH1 was frequently mutated in diffuse astrocytoma (68-88%), oligodendrogliomas (69-79%) and oligoastrocytomas (78-94%) as well as secondary glioblastomas (88%) (Balss et al., 2008; Watanabe et al., 2009). In 2009, Yan et al. were the first to sequence IDH2 in brain tumors and found that IDH mutations were not confined to IDH1 but were also found in IDH2. The lysine that was found to be mutated in IDH2 was R172 and is in the homologous position to IDH1 R132 (Yan et al., 2009)

One of the first studies to describe a functional consequence of mutant IDH1 and IDH2 was in relationship to hypoxia-inducible factor 1 (HIF-1). At the time it was thought that mutations in IDH1/2 were simple loss of function mutations, resulting in less KG production. KG is an essential co-factor for prolyl hydroxylases that promote the degradation of HIF-1. When IDH1 or IDH2 are mutated less KG is available to bind to prolyl hydroxylases and thus 1 levels increase, this in turn leads to an increase in HIF-1 dependent genes that are involved with numerous pathways needed for tumor growth (Zhao et al., 2009).

(37)

20 Mutant IDH1 produces 2HG

In a landmark study by Dang et al. they discovered that mutations in IDH1 are not simple loss of function mutations but are in fact gain of function or neomorphic mutations. mutant IDH1 does not just lose the ability to produce KG; instead IDH1 gains the ability to reduce KG and produce 2-hydroxyglutarte (2HG) in an NADPH dependent manner (Figure 1.2). These results provided a rationale as to why IDH mutations are always heterozygous: WT and mutant IDH1 heterodimerize and the products of WT IDH1, KG and NADPH, became the substrates for mutant IDH1 to produce 2HG and NADP+. No significant differences in the ability to catalyze this new reaction were found with different amino acid substitutions, showing that reduction of KG to 2HG is a universal function of mutant IDH1. When analyzed across a panel of patient brain tumor samples, 2HG levels were significantly increased in mutant IDH1 tumors compared to tumors with WT IDH1 (Dang et al., 2009). IDH1/2 mutations in AML

A year after IDH1/2 mutations were found in brain tumors, similar mutations were identified in AML. In a study of 188 AML patients, 16 (8.5%) had mutations in IDH1. In contrast to brain tumors that predominately have R132H substitutions, a more diverse group of substitutions were detected in AML; 8 had R132C, 7 had R132H and 1 had R132S (Mardis et al., 2009). Following, a large-scale study of 350 AML patients revealed that not only did AML patients harbor IDH1 mutations, but also IDH2 mutations. In addition to the IDH2 R172 mutation described in brain tumors, a novel IDH2 mutation at R140 was also detected. Overall, 25% of AML patients have either an IDH1 or IDH2 mutation (Marcucci et al., 2010; Papaemmanuil et al., 2016).

(38)

21

Upon finding that IDH2 mutations also existed in AML patients, Ward et al. demonstrated that 2HG production is a common feature in AML patients. Indeed, they clearly showed that IDH2 R172K mutants also produces high levels of 2HG in a NADPH dependent manner. In an effort to determine if 2HG could be used as a biomarker to detect IDH1/2 mutations they confirmed the existence of a second mutated arginine in IDH2, R140Q, which is also found in the active site and is able to produce 2HG. It was found that in all patients where 2HG was detected, an IDH1/2 mutation was also detected, providing strong evidence that 2HG can be used as a biomarker (Ward et al., 2010). Upon the establishment that both mutant IDH1/2 produced 2HG in AML and gliomas, multiple groups began in depth investigations to understand the functional role of 2HG.

Biological effects of 2HG

Multiple enzymes rely on KG; therefore, multiple enzymes can be inhibited by 2HG. The first large scale AML study investigating the effects of 2HG linked mutant IDH1/2 and DNA methylation. By clustering differentially methylated regions in over 400 AML patients the authors revealed that mutant IDH1/2 shared a common DNA methylation profile that was distinct from other leukemic drivers. Interestingly, when the differentially methylated regions and subsequently downregulated genes in mutant IDH1/2 samples were compared to those in TET2 mutant samples, a significant overlap was found. TET2 and IDH1/2 mutants were found to be mutually exclusive, strongly suggesting they affect similar pathways. To test this hypothesis they showed that 2HG produced by mutant IDH1/2 could directly increase 5mC, very likely through the inhibition of TET2. In in vitro assays expression of mutant IDH1/2 or knockdown of TET2 each lead to a block in differentiation (Figueroa et al., 2010a). The first IDH1 and IDH2 inhibitors successfully decreased 2HG levels and promoted

(39)

22

differentiation in both glioma and leukemia cells, respectively (Rohle et al., 2013; Wang et al., 2013).

While most studies investigating mutant IDH1/2 focus on understanding the effects of 2HG on a single enzyme, Xu et al systematically documented changes induced by 2HG in every group of KG-dependent enzyme known at the time (Xu et al., 2011a). The significance of this work lies in the fact that it clearly demonstrates the incredibly diverse effects of IDH mutations on the cell, and that it is not just one enzyme that drives leukemogenesis. 2HG inhibits histone demethylases, globally increasing histone methylation, and altering gene expression; 2HG inhibits the prolyl hydroxylase responsible for regulating HIF-1, resulting in increased levels of HIF-1, and changes in the downstream targets; and 2HG inhibits TET2 leading to decreased levels of 5hmC (Figure 1.3). Perhaps most interestingly though was the biochemical work, which revealed 2HG to be a weak competitive inhibitor and as much as 100-fold increase of 2HG must be available to outcompete KG (Xu et al., 2011a). This is important evidence that no cancer with an IDH1 or IDH2 mutation will be exactly the same and most likely all of the enzymes found to be inhibited by 2HG are in fact inhibited.

Leukemic co-drivers with mutant IDH1/2

Mutant IDH1/2 are unable to drive leukemia alone. This has been shown both in mouse models and in humans. Mutations in IDH1/2 are often early mutations and may encourage myelodysplastic syndrome (MDS) but need at least one other mutation to progress to AML. The top co-occurring mutations with mutant IDH1/2 are NPM1c, FLT3-ITD and MLL-PTD (Cancer Genome Atlas Research et al., 2013; Patel et al., 2012). In mouse models of mutant IDH2 and FLT3-ITD neither are able to produce leukemia but when combined result in an aggressive leukemia (Kats et al., 2014).

(40)

23 Figure 1.3: Enzymes inhibited by 2HG

2HG, a structurally similar molecule of KG, competitively inhibits the many enzymes that require KG as a co-factor.

(41)

24

Two significant mouse models involving mutant IDH1/2 are those with retroviral overexpression of HOXA9 or HOXA9 and MEIS1. Overexpression of HOXA9 in a stem cell enriched population of mouse bone marrow cells is sufficient to produce leukemia. However, it was found that when mutant IDH1 was co-transduced with HOXA9 transplanted mice died of leukemia faster (Chaturvedi et al., 2013). These results strongly suggesting a key role for mutant IDH1 in leukemia. Following, it was found that when mouse bone marrow cells from an IDH2 knock-in mouse were transduced with HOXA9/MEIS1 a dependency on mutant IDH2 developed. In this system the mouse had been accumulating 2HG in its bone marrow for most of its life. When mutant IDH2 was then removed from the developing leukemia the cells could no longer serially re-transplant (Kats et al., 2014). HOXA9 and MEIS1 are key downstream targets of MLL-r; therefore, these results led us to question if mutant IDH1/2 would cooperate with MLL-r.

We find in chapter II, through the use of multiple mouse models, no cooperation between MLL-r and mutant IDH1/2 exists. In fact, mutant IDH1/2 have detrimental effects on MLL-r AML and can downregulate key MLL-r target genes. By overexpressing DOT1L in chapter III we find that neither increases nor decreases in H3K79me2 are tolerated in MLL-F AML and both rapidly result in a downregulation of the MLL-r program. Decreased gene expression is the result of a loss of a differential in H3K79me2 levels between target and non-target genes that exists because MLL-F have in fact low levels of H3K79me2 everywhere but the target genes. On the other hand, we also find in chapter IV that mutant IDH1/2 and NPM1c leukemia have significantly high global levels of H3K79me2. Finally, we provide evidence that mutant IDH1/2 can independently induce expression of the HOX genes.

(42)

25

CHAPTER II

MUTANT IDH1/2 AND MLL-REARRANGEMENTS DO NOT COOPERATE TO DRIVE

LEUKEMOGENESIS

Abstract

Mutations in isocitrate dehydrogenase 1 and 2 (IDH1/2), commonly found in adult leukemias, are capable of dysregulating RNA, DNA, and histone methylation, as well as metabolic and hypoxia pathways. A targeted inhibitor has recently been approved by the FDA, making the need to fully understand the effects brought on by the mutant protein more important than ever. Partial tandem duplications of the mixed lineage leukemia (MLL-PTD) protein, a type of MLL-rearrangement (MLL-r), are frequently found with mutations in IDH1/2. Using murine models and human cell lines, we sought to understand if mutant IDH1/2 contributed to leukemogenesis when in combination with MLL-r. We found no cooperation and, in some cases, impaired growth, that was a result of a downregulation of genes in the HOX cluster.

Introduction

Mutations in isocitrate dehydrogenase 1 and 2 (IDH1/2) are found in 20-30% of adult acute myeloid leukemia (AML) patients (Cancer Genome Atlas Research et al., 2013; Patel et al., 2012). Wild type IDH1/2 produce -ketoglutarate (KG), an essential co-factor for numerous enzymes including DNA, histone, and RNA demethylases, collagen prolyl hydroxylases, and HIF hydroxylases (Chowdhury et al., 2011; Figueroa et al., 2010a; Koivunen et al., 2012; Rose et al., 2011; Sasaki et al., 2012a; Su et al., 2017; Xu et al., 2011a;

(43)

26

Zhao et al., 2009). Single point mutations in the enzymatic region of IDH1/2 confer the ability to reduce KG to 2-hydroxyglutarate (2HG), a competitive inhibitor of KG (Dang et al., 2009; Ward et al., 2010). Thus, mutations in IDH1/2 have the potential to lead to a multitude of different epigenetic alterations. Mutant IDH1/2 are often founding mutations, appearing early in the clonal evolution of AML. However, an IDH1/2 mutation is not enough to drive frank leukemia alone; additional co-occurring mutations are necessary for leukemogenesis (Bowman et al., 2018; Sperling et al., 2016; Steensma et al., 2015). The top five most common co-occurring mutations are NPM1c, DNMT3A, FLT3-ITD, RUNX1, and MLL-PTD (Cancer Genome Atlas Research et al., 2013; Patel et al., 2012).

One biological change observed upon introduction of mutant IDH1/2 or 2HG exposure found by three different groups was an increase in H3K79me2. The increase in H3K79me2 was found in glioma cells treated with 2HG (Xu et al., 2011a), 3T3 cells with retroviral overexpression of mutant IDH1 or mutant IDH2 (Lu et al., 2012), and in the bone marrow of IDH1 knock-in mice (Sasaki et al., 2012b). Due to the fact that multiple subtypes of leukemia are dependent on H3K79me2 we found the increase in H3K79me2 brought on by mutant IDH1/2 particularly interesting. In many of the leukemias that are dependent on H3K79me2, genes in the HOX cluster are highly expressed (Bernt et al., 2011; Daigle et al., 2011; Kuhn et al., 2016; Ku hn et al., 2015b). In the glioma model, the increase in H3K79me2 was concordant with an increase in expression of genes in the HOXA cluster (Xu et al., 2011a). Additionally, in a study of AML patients with mutations in IDH1/2 genes in the HOXA cluster were highly expressed (Chaturvedi et al., 2013). The findings that mutations in IDH1/2 increased H3K79me2 levels and had high HOX gene expression, led to the question of a potential cooperation between mutant IDH1 and HOXA9. Indeed, it was found that retroviral

(44)

27

expression of both mIDH1 and HOXA9 in a murine bone marrow accelerated the development of leukemia versus overexpression of HOXA9 alone. (Chaturvedi et al., 2013). Even more interestingly, when HOXA9/MEIS driven leukemia was established in a mutant IDH2 knock-in mouse, a dependency on mutant IDH2 developed as de-induction of mutant IDH2, using a TET on system, resulted in a significant reduction in growth in vitro with failure to propagate leukemia in secondary recipients. (Kats et al., 2014).

High expression of genes in the HOXA cluster is a key hallmark of all leukemias driven by rearrangements of the mixed lineage leukemia gene (r). r encompass both MLL-Fusions (MLL-F) and MLL-partial tandem duplications (MLL-PTD). In MLL-F, the C-terminal end of the protein is lost while the N-terminal is fused to one of over 100 different partners. In MLL-PTD, the protein remains whole, but exons 2 through 8-11 are duplicated. While MLL-F are strong oncogenic drivers that are capable of driving leukemia alone, MLL-PTD always require additional mutations for full leukemic transformation (Andersson et al., 2015; Krivtsov and Armstrong, 2007; Sun et al., 2016). Interestingly, the two three most commonly co-occurring mutations with MLL-PTD are FLT3-ITD, mutant IDH1/2, and DNMT3a (Sun et al., 2016).

To date, all MLL-r leukemias tested have been found to be dependent on the H3K79 methyltransferase, DOT1L, and H3K79me2 on selected genes, specifically the HOXA cluster (Bernt et al., 2011; Deshpande et al., 2013; Ku hn et al., 2015a). It was recently shown that mutant IDH1/2 leukemias were also dependent on DOT1L (Sarkaria et al., 2014). Given that mutant IDH1/2 were found to cooperate with HOXA9/MEIS1, and have been shown to cause increased H3K79me2 (Lu et al., 2012; Ogawara et al., 2015; Sasaki et al., 2012b; Xu et al., 2011a), we set out to determine whether mutant IDH1/2 could cooperate with MLL-F

(45)

28

and/or MLL-PTD. We hypothesized that the addition of mutant IDH1/2 could result in stabilization or increase of H3K79me2 on MLL-r targets, such as HOXA9, thus directly contribute to leukemogenesis.

Methods Virus production

Ectopic retroviral vectors (10 µg) containing murine MLL-AF9-IRES-YFP (MLL-AF9-YFP), MLL-AF9-IRES-GFP (MLL-AF9-GFP), IDH1 R132H-IRES-GFP (IDH1R132H), MSCV-IRES-dTomato (MIT), or MSCV-IRES-GFP (MIG) were co-transfected with delta psi packaging plasmid (10 µg) in 293T cells using Fugene® and Optimem. Alternatively, VSVG pseudotyped lentiviral vectors (10 µg) containing IDH1R132C-IRES-GFP (IDH1R132C) or pHIV-IRES-dTomato (PIT) were contransfected with pΔ8.9 (1 µg) and VSVG (9 µg) packaging plasmids in 293T cells using Fugene® and Optimem. Supernatant containing viral particles was collected 12, 36 and 72 hours after transfection. To concentrate the virus, polyethylene glycol was added to viral containing media at a ratio of 1:5 overnight at 4 °C, followed by centrifugation at 2500 rpm for 20 minutes at 4 °C. The concentrated virus was resuspended in PBS at a 1:100 ratio.

Cell culture

The following cell lines were purchased from the Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ, Germany) or the American Type Culture Collection (ATCC, USA): EOL-1, THP1, RS4;11, Kasumi, 293T, and 3T3.

EOL-1, THP1, RS4;11, and Kasumi cells were maintained in RPMI-1640 media supplemented with heat-inactivated 10% fetal bovine serum and 50 U/ml Penicillin/Streptomycin. 293T and 3T3 cells were maintained in DMEM supplemented with

(46)

29

heat-inactivated 10% fetal bovine serum and 50 U/ml Penicillin/Streptomycin. The patient sample (2614; IDH1R132S/FLT3-ITD) was cultured in RPMI-1640 supplemented with 10% FBS and 50 U/mL penicillin/streptomycin. Generated murine leukemias were maintained in IMDM supplemented with 15% FBS, 50 U/ml Penicillin/Streptomycin, 1% L-glutamine, and 10 ng/ml murine IL3 and IL6, and 20 ng/ml murine SCF or M3234 methylcellulose supplemented with 50 U/ml Penicillin/Streptomycin, 10 ng/ml murine IL3 and IL6, and 20 ng/ml murine SCF, and IMDM. All cells were cultured in the stated media in a humidified incubator at 37°C in 5% CO2.

Generation of murine leukemia

Bone marrow cell suspensions from wild type mice, IDH2R140Q knock-in (kindly

provided by Craig Thompson at Memorial Sloan Kettering Cancer Institute), or Vav1-Cre (Jax number: 018968) litter mate controls were prepared by crushing arm, leg, and hip bones in a mortar after removal of muscle and connective tissues. The crushed bones were filtered through a 0.2-micron filter and spun at 1200 rpm for 5 minutes. The supernatant was removed, and red blood cells were lysed on ice using red blood cells lysis buffer Pharm Lyse. Lineage depletion was performed by labeling bone marrow cell suspensions with a mixture of purified biotinylated monoclonal antibodies to CD3e, CD4, CD8a, CD19, B220, Gr-1, IL-7R and Ter-119. Cell positive for any of the lineage markers were partially removed by two rounds of magnetic bead depletion with streptavidin conjugated Dynabeads. The lineage depleted (lin-) cells were stained with streptavidin (APC-Cy7), c-Kit (Alexa 647), Sca-1 (PE-Cy7) and sorted for Lin- Sca-1+ cKit+ (LSK). Sorted cells were pre-stimulated for 24 h with 10 ng/ml murine IL3 and IL6 and 20 ng/ml murine SCF, Flt3L and TPO. Transduction of LSK cells was carried out on retronectin coated plates with MLL-AF9 and IDH1R132H viruses

(47)

30

(when indicated) in the presence of murine IL3, IL6, SCF, Flt3L, and TPO in concentrations stated above. Cells were subsequently maintained in M3234 methylcellulose supplemented with murine IL3, IL6, SCF (at the above concentrations), 50 U/ml Penicillin/Streptomycin and IMDM. After 2-3 days, GFP+, YFP+ or GFP+/YFP+ cells were sorted and either transplanted into mice or maintained in methylcellulose. 6-week-old C57BL/6 female mice were sub-lethally irradiated (600 or 550 RAD) and transplanted with the number of cells indicated in the text.

Transplanted mice were monitored daily for signs of leukemic progression which includes a hunched posture, slowed movement, dehydration, weight loss, and/or changes in grooming patterns. Twenty-one days after transplant, mice were subject to retro-orbital bleeds to assess engraftment in the peripheral blood. When mice began to exhibit signs of leukemia, they were euthanized using CO2 followed by cervical dislocation. Necropsy analysis included: a heart puncture to collect blood for complete blood counts (CBC), removal of the spleen, liver, spine, lymph nodes (if enlarged), and thymus (if enlarged) into 10% formaldehyde, and removal of the leg bones. Leg bones were crushed, the red blood cells lysed, and the cells analyzed for percentage of GFP and/or YFP by flow cytometry. Once confirming there was at least 90% leukemic cells in the bone marrow the cells were frozen or plated directly into methylcellulose or liquid culture.

Colony formation assay

1,000 or 500 in vivo transformed cells were plated in M3234 methylcellulose with murine IL3, IL6, SCF, 50 U/ml Penicillin/Streptomycin and IMDM. Every 5-7 days the colonies were scored based upon morphology, total cell numbers counted, and replated in fresh methylcellulose.

(48)

31 Cell growth and drug assays

The DOT1L inhibitor, EPZ5676, IDH1 inhibitor, AGI-5198, IDH2 inhibitor, AGI-6780, or DMSO control was added at the indicated concentrations. Cell growth and viability were determined by trypan blue exclusion staining every 3 -4 days and cells were replated in fresh media with fresh drug at equal densities for 10-15 days.

Competition assay

EOL-1 cells were transduced with IDH1R132C or PIT, and MLL-AF9 murine leukemia cells were transduced with IDH1R132H or MIT on retronectin coated plates. Forty-eight hours after transduction GFP+ or dTomato+ cells were sorted on the MoFlo Astrios Flow Sorter. Non-transduced cells were also sorted on singlets. Sorted cells were counted by trypan blue exclusion staining and the indicated populations were plated together in a 1:1 ratio. To verify starting ratio, a small amount of the plated cells was taken and analyzed using the CytoFLEX Flow Cytometer (day 0). Every 3 days, cells were counted by trypan blue exclusion staining, a set number was replated in fresh media, and the rest was analyzed on the CytoFLEX Flow Cytometer to determine the ration of the two cell populations in the well.

2HG treatment

Cell permeable 2HG was dissolved in PBS to a concentration of 50 mM. 2HG was added at the indicated concentration. After 48 hours cells were counted by trypan blue exclusion staining and replated in fresh media with fresh 2HG. At 96 hours final counts were performed by trypan blue exclusion staining or an XTT colometetric assay.

Histone and protein isolation and western blots

Histones were extracted by lysing cell pellets in 0.5% tritonX in PBS for 15 minutes on ice. Nuclei were pelleted for 10 minutes at 15000 rpm in a standard benchtop centrifuge

(49)

32

at 4°C. The nuclei were washed in 0.5% tritonX in PBS and centrifuged for 10 minutes at 15000 rpm in a standard benchtop centrifuge at 4 °C. The histones were then extracted overnight at 4°C in 0.2N HCl. For long-term storage histones were kept at -20°C.

To extract whole protein lysates, cells were lysed in RIPA (150mM sodium chloride, 50mM Tris pH 8, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS) buffer with protease inhibitor for 15 minutes on ice. The lysed cells were centrifuged for 10 minutes at 15000 rpm in a standard benchtop centrifuge at 4°C. The supernatant was collected, added to 4X LDS buffer and 10X reducing agent, and boiled at 85°C for 5 minutes. For long-term storage, whole protein lysates were kept at -20°C.

Whole protein lysates or histones were separated on a 10% Bis-Tris gel, blotted on nitrocellulose membrane, and blocked in 5% non-fat milk made with TBS-T for 1 hour. All antibodies were diluted into 5% non-fat milk made with TBS-T. Total H3, total H4 and actin were used at 1:2000, all other primary antibodies were used at 1:1000. Secondary antibodies were used at 1:10000. Proteins were visualized using Western Lightning Plus-ECL.

RNA extraction

Cell pellets were resuspended in Qiagen RLT buffer with 2-Mercaptoethanol (1:100) and frozen at -80 °C. For extraction of RNA, samples were warmed to room temperature and extracted using the RNeasy kit (Mini kit for over 500,000 cells and micro kit for under 500,000 cells). Briefly, gDNA was removed by a column, the flow through was mixed 1:1 with 70% ethanol, the RNA was bound to a second column and washed with ethanol. RNA was eluted in water and quantified using a NanoDrop Spectophotometer.

References

Related documents

AKN-028 is a potent inhibitor of FLT3 with an IC 50 value of 6 nM in an enzyme assay, but also displaying in vitro activity in a variety of primary AML samples, irrespective of

Mechanistic exploration was performed using the NextBio bioinformatic software using gene expression analysis of drug exposed acute myeloid leukemia cultures (HL-60) in the

Increased prevalence of prior malignancies and autoimmune diseases in patients diagnosed with chronic myeloid leukemia.. Second malignancies following treatment of

For paper I the Swedish cancer register (SCR) was used to identify all Swedish CML patients since 1970 and the Swedish cause of death register to identify an eventual

The addition of HDC or other ROS inhibitors or scavengers rescued a significant proportion of NK cells (Figure 8B). These results support the hypothesis that

However, NK cells are implicated significant effector cells in several malignancies, including acute myeloid leukemia where the leukemic blasts mostly express HLA class I Paper I,

Additionally, a single nucleotide polymorphism in HLA-B that dictates NK cell inhibition to be preferentially mediated by NKG2A impacted positively on outcome in this trial

The aims were (I) to describe a European lower risk MDS population and the use of erythropoietin stimulation agents (ESA), (II) to describe the AML population in