Research Collection

Doctoral Thesis

HIF1α dependant transcriptional networks in macrophages and hepatocytes

Author(s): Müller, Julius

Publication Date: 2009

Permanent Link: https://doi.org/10.3929/ethz-a-005900145

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For more information please consult the Terms of use.

ETH Library

DISS. ETH Nr. 18594

HIF1 ααα dependant transcriptional networks in macrophages and hepatocytes

ABHANDLUNG zur Erlangung des Titels

DOKTOR DER WISSENSCHAFTEN der ETH ZÜRICH

vorgelegt von Julius Müller

Dipl. Biol., Ruprecht-Karls-Universität Heidelberg

geboren am 16.09.1977

von Deutschland

Angenommen auf Antrag von Prof. Romeo Ricci Prof. Wilhelm Krek Prof. Peter Bühlmann

2009

Index

1. Index

1. Index ...... 2

2. Acknowledgements ...... 4

3. Summary ...... 5

4. Zusammenfassung ...... 6

5. Abbreviations ...... 7

6. Introduction ...... 8 6.1. Hypoxia and the Hypoxia Inducible Factor 1 alpha ...... 8 6.2. Transcriptional- and epigenetic regulation ...... 15 6.3. Methodology to address Genome wide binding patterns...... 22

7. Aim of the project ...... 24

8. Material and Methods ...... 25 8.1. Media and Buffers used for ChIP and ChIP-chip ...... 25 8.2. Cell lines ...... 28 8.3. ChIP-on-chip...... 29 8.4. ChIP-Seq ...... 35 8.5. mRNA Expression Profiling ...... 35 8.6. Identification of ChIP-chip Peaks ...... 36 8.7. Identification of ChIP-Seq Peaks ...... 37 8.8. De novo Motif Analysis ...... 37 8.9. Q-PCR Validation of ChIP Hits ...... 38 8.10. Annotation of sequences and association of expression- to binding data ...... 38

9. Results ...... 39 9.1. Regulation of HIF1 α and its target genes ...... 39 9.2. Genome wide expression and binding studies ...... 42 9.3. Genome wide binding study using the murine leukemic monocyte-macrophage cell line (Raw.264) and ChIP-Seq ...... 54 9.4. Promoters that are occupied by HIF1 α in Raw.264 cells ...... 57 9.5. Characterization of the Hypoxia Response Element ...... 63 9.6. Transcription factors interacting with Hif1 α ...... 68 9.7. Downstream regulatory mechanism regulated by Hif1 α ...... 72

Page 2 Index

10. Discussion...... 77 10.1. Binding of HIF1 α is cell type specific ...... 77 10.2. One out of five genes that are bound by HIF1 α are differentially expressed in PMH and PMM 78 10.3. One out of twenty-five hypoxia responsive genes are bound by HIF1 α in PMH ...... 79 10.4. ChIP-Seq reveals markedly more HIF1 α binding events during hypoxia in Raw.264 cells .. 80 10.5. HIF1 α directly binds to genes associated to glycolysis, angiogenesis and regulation of transcription, depending on the cell type...... 81 10.6. HIF1 α preferentially binds close to the TSS ...... 81 10.7. HIF1 α preferentially binds to an HRE consisting of nine base pairs or an tandem core HRE 82 10.8. The TRE consensus motif is overrepresented at enhancer regions targeted by HIF1 α ...... 83 10.9. SP1 is a potential HIF1 α target and might regulate genes in response to hypoxia independent of HIF1 α ...... 84 10.10. Transcriptional regulation of chromatin modifiers by HIF1 α ...... 85 10.11. Comparison to previous genome wide HIF1 α binding studies...... 86

11. Outlook ...... 88

12. References ...... 90

13. Supplements ...... 97 13.1. Top 300 up regulated genes in PMH and PMM ...... 97 13.2. Group III genes of PMM, PMH and Raw.264 cells (Top 300) ...... 104

Page 3 Acknowledgements

2. Acknowledgements

Ich möchte meine Arbeit den folgenden Personen widmen, die alle direkt und indirekt zum erfolgreichen Abschluss meiner Doktorabeit während der letzten vier Jahre beigetragen haben:

Meinen Eltern, ohne dessen Unterstützung diese Arbeit niemals möglich gewesen wäre.

Romeo Ricci, der mir die Möglichkeit gegeben hat, dieses Projekt bis zum Ende durchzuführen.

Meinen Kollegen: Renata Windak, Grzegorz Sumara, Susann Kumpf, Arne Ittner, Helmuth Gehart und Ivan Formentini

Meine Semesterstudenten: Yvonne Fink und Andreas Essig

Meinem Thesis Committee: Wilhelm Krek und Peter Bühlmann

Meinen Kollaboratoren: Andrea Patrignagni (ChIP-chip), Bernard Jost (ChIP-Seq)

Meinen Geschwistern: Caroline, Wiebke und Oskar

Weitere wichtige Elemente: Susann (Administration und Lehre), Arne und Helmuth (Abend Snacks), Nikolai (Kaffeepausen), Felix (Trainingspartner), Antra (Rauchen), Gerald (Ernährung), Stefan und Strahil (Golf Pros), Iza (General Support), Ivan (Fluchen)…

Page 4 Summary

3. Summary

HIF1 α is the principal transcription factor that mediates responses to low oxygen levels in eukaryotic cells. By comparing genome-wide promoter binding studies of HIF1 α in primary mouse hepatocytes and primary mouse macrophages, I was able to demonstrate that HIF1 α binding is cell-type specific. Integration of expression data revealed that only a small fraction of genes bound by HIF1 α are differentially expressed. To explore the transcriptional mechanisms, which modulate the differentially expressed genes secondary or independent of HIF1 α, motif analysis of the respective promoters was performed. Among others, transcription factors of the activating protein 2 (AP2) family and SP1 showed a marked overrepresentation suggesting an important function in the regulation of hypoxia responsive genes. To complement these results with genome-wide binding data, an unbiased binding study of HIF1 α using ChIP-Seq in Raw.264 cells, was performed. Although the majority of binding events were localized close to the transcriptional start side (TSS), about 40% of the peaks occurred more than 10kbp away from the TSS. Motif analysis of the Raw.264 binding study revealed that HIF1 α preferably binds to an extended hypoxia response element (HRE) and in 14% of the cases, a tandem core HRE seems to be the consensus site bound by HIF1 α. Moreover, I showed that AP1 might be an important factor cooperating with HIF1 α at enhancer sites to modulate expression levels of developmental genes and genes associated to apoptosis. Another mechanism of transcriptional regulation upon hypoxia may involve the JmjC family of histone demethylases that were found to be direct targets of HIF1 α. Thus, my data refined the HRE motif that appears to be bound by HIF1 α in a cell- specific manner. Furthermore, this work demonstrates that the hypoxia response in mammalian cells is to a larger extent regulated by other transcription factors and by dynamic epigenetic changes both of which may be dependent and independent of HIF1 α.

Page 5 Zusammenfassung

4. Zusammenfassung

Sauerstoffmangel oder Hypoxie wird von allen eukaryotischen Zellen hauptsächlich durch den Transkriptionsfaktor HIF1 α in eine transkriptionelle Antwort übersetzt. In dieser Arbeit wurde durch Genom-weite Promoter-Bindungsanalysen dieses Transkriptionsfaktors gezeigt, daß HIF1 α zelltypspezifisch an die DNA bindet. Außerdem konnte gezeigt werden, daß nur ein kleiner Teil der HIF1 α gebunden Gene auch eine Änderung der Transkript Konzentrationen erfährt. Um die großen Unterschiede zwischen der Anzahl der deregulierten und der gebundenen Gene zu erklären, wurde eine de novo Promoter Sequenzanalyse der deregulierten, aber nicht gebundenen Gene durchgeführt. Hier zeigte sich, daß die klassischen Erkennungssequenzen Transkriptionsfaktoren SP1 und AP2 signifikant überrepräsentiert waren. Weiterhin wurden die Daten durch ChIP-Seq Daten einer murinen Makrophagen Krebszelllinie ergänzt. Hier zeigte sich das trotz einer starken Konzentration der Bindungsstellen am Promoter, etwa 40% der Bindungsstellen mehr als 10tbp entfernt lokalisiert waren. Weiterhin konnte durch eine de novo Motivanalyse der gebundenen Sequenzen ein Tandem Erkennungsmotiv von HIF1 α, das HRE, in 14% der gebundenen Sequenzen, gefunden werden. Interessanterweise konnten zusätzlich Transkriptionsfaktoren der AP1 Familie mit etwa 16% der gebundenen Sequenzen in Verbindung gebracht werden. Diese Bindungsstellen waren weiter vom Transkriptionsstart entfernt und es kann spekuliert werden, daß AP1 Transkriptionsfaktoren wichtig für die Co-Regulation von HIF1 α gebundenen Genen der zellulären Entwicklung und der Apoptose ist. Ein weiterer Mechanismus der induzierten, transkriptionellen Kontrolle durch Hypoxie, können Histon-demethylasen der JmjC Familie sein, die direkt von HIF1 α gebunden und dereguliert werden. Zusammenfassend kann gesagt werden, daß HIF1 α zelltypspezifisch an sein Erkennungsmotiv bindet. Weiterhin zeigt diese Arbeit, daß die transkriptionelle Antwort auf Hypoxie zu einem Großteil auf andere Transkriptionsfaktoren und epigenetischen Mechanismen beruht, die direkt oder indirekt durch HIF1 α moduliert werden.

Page 6 Abbreviations

5. Abbreviations

CDS Coding Sequence ChIP Ch romatin-Immuno precipitation ChIP-chip Analysis of the ChIPed DNA by DNA-microarrays ChIP-Seq Analysis of the ChIPed DNA by Seq uencing DAVID Database for Annotation, Visualization and Integrated Discovery FDR False discovery rate GO Gene Ontology HRE Hypoxia Response Element LPS Lipo poly saccharide, an endotoxin that triggers an inflammatory response LSC Location and Size matched, randomized Control region PMH Primary, Liver perfusion elicited Mouse Hepatocytes PMM Primary, Thioglycollate elicited Mouse Macrophages PP Proximal Promoter -> -5000 to +3000kbp to the TSS TRE TPA Response Element TSS Transcriptional Start Site WCE Whole Cell Extract

Page 7 Introduction

6. Introduction

6.1. Hypoxia and the Hypoxia Inducible Factor 1 alpha

6.1.1. Cellular oxygen demand and hypoxia

Molecular oxygen is indispensable to maintain the normal physiological status of all mammalian cells. Most importantly, oxidative phosphorylation in mitochondria is dependent on molecular oxygen in order to provide cells with sufficient energy. Non- mitochondrial oxygen consumption is accounting for up to 10–30% of total cellular O 2 consumption (Herst and Berridge, 2007; Rosenfeld et al., 2002). The physiological oxygen levels to accomplish these needs can span a wide range. Assuming atmospheric oxygen partial pressure of 21 kPa (equals 21%), blood oxygen levels are about 13 kPa in the arterial and 9 kPa in the venous blood, respectively. Due to the different diffusion and vascularization conditions, oxygen levels described for tissues can only be roughly estimated. For example, in vivo measurements of partial pressure in mouse spleens revealed values of about 0.5–4.5 kPa, depending on the distance from the artery (Caldwell et al., 2001). However, if oxygen supply becomes insufficient, the physiological status of the cell can undergo drastic changes. Changes in cell physiology include pH status (Chiche et al., 2009), the abundance of reactive oxygen species (ROS) (Guzy and Schumacker, 2006), genome integrity (To et al., 2006), growth and cell survival (Carmeliet et al., 1998; Liu et al., 2006), protein translation (Young et al., 2008) and iron metabolism (Peyssonnaux et al., 2008). In fact, prolonged hypoxia inevitably leads to cell death. Therefore, sophisticated systems evolved to provide metazoan organisms with sufficient oxygen such as the cardiovascular system and the respiratory system. Moreover, all eukaryotic cells possess complex mechanisms to sense and adapt to low oxygen levels (hypoxia) and many of these adaptation processes, if deregulated, can play an important role in the development and progression of a variety of diseases including atherosclerosis (Sluimer and Daemen, 2009), cancer (Denko, 2008), diabetes (Crawford et al., 2009) and inflammatory disorders (Sitkovsky and Lukashev, 2005).

Page 8 Introduction

6.1.2. HIF1 α, the principal mediator of the hypoxic response in mammals

A central mediator of the hypoxic response in all mammalian cells is the transcription factor hypoxia inducible factor 1 alpha (HIF1 α) (Iyer et al., 1998). It belongs to the family of the Hypoxia Inducible Factors, and is the only isoform that is ubiquitously expressed. HIF1 α is tightly regulated by oxygen levels. Under normoxic conditions, HIF1 α is targeted by one or more of the three prolyl hydroxylase domain proteins (PHD1–3) which can hydroxylate the oxygen dependant degradation domain (ODDD) on two different Proline residues (Kaelin and Ratcliffe, 2008; Schofield and Ratcliffe, 2004). This hydroxylation leads to rapid recognition and degradation of HIF1 α that is mediated by VHL, a component of an E3 multiprotein ubiquitin-ligase complex (Figure 1A). Since PHD enzymes belong to the family of highly oxygen-dependent 2-oxoglutarate- dependent-oxygenases, HIF1 α is stabilized under hypoxia, binds to different importins and gets translocated to the nucleus where it heterodimerizes with constitutively expressed HIF1 β (Depping et al., 2008). The dimerisation between the HIF1 α and HIF1 β subunits occurs through the basic helix-loop-helix (bHLH) and PER-ARNT-SIM (PAS) A and B domains located in the N-terminal region of each subunit, whereas DNA binding to the hypoxia response element (HRE), occurs through the bHLH domains (Brahimi- Horn and Pouyssegur, 2009). However, to give rise to a transcriptional response, the presence and interaction of other co-activators such as CBP/p300 is required. This interaction can be prevented by the factor inhibiting HIF (FIH), which can hydroxylate asparagine residue within the carboxy terminal transcriptional activation domain (CTD) (Lando et al., 2002). Since FIH also belongs to the family of 2-oxoglutarate-dependent- oxygenases, asparagine hydroxylation represents a second, oxygen-dependent mechanism to regulate HIF1 α activity.

Page 9 Introduction

6.1.3. Alternative stabilization and regulation of HIF1α

HIF regulation is not limited to low oxygen levels. An overview about the current knowledge of hypoxia-independent regulatory mechanism of HIF1 α is depicted in Figure 1B, which was taken from a recent review (Brahimi-Horn and Pouyssegur, 2009). It depicts the broad variety of possibilities to regulate HIF1 α independent of hypoxia and therefore underlines its importance in the normal physiology of the cell.

Figure 1: Hypoxia -dependent regulation and hypoxia -independent activation of HIF. (A) The scheme depicts the oxygen-depend ent degradation of different HIF family members. In brief, under normoxia HIFs can be hydroxylated by PHDs or FIH, which leads to proteasomal degradation or inability to interact with obligatory co activators (CBP/p300) respectively. Adopted from (Schofield and Ratcliffe, 2004) (B) The scheme depicts the most well established means o f hypoxia-independent regulation of HIF1 α on the transcriptional level, on the translational level, on the posttranslational level and on the level of transcriptional activity of HIF1 α. Adopted from (Brahimi-Horn and Pouyssegur, 2009).

Page 10 Introduction

For example, on the transcriptional level, HIF1 α mRNA expression can be enhanced by LPS exposure in macrophages, presumably through a NF-κB-mediated mechanism (Belaiba et al., 2007; Frede et al., 2006). Also on the translational level, HIF1 α protein can be regulated by enhanced translation via the mTOR/Akt pathway (Harada et al., 2009), and, less understood, specifically by different proposed mechanisms involving e.g. CAP-independent translation via its internal ribosomal entry side (IRS) (Yee Koh et al., 2008). Additionally, HIF1 α transcriptional activity can be regulated by a variety of modifications (Lisy and Peet, 2008). The best-understood modifications, apart from hydroxylation of asparagines by FIH described above, include mitogen-mediated phosphorylation (Richard et al., 1999), acetylation of a lysine residue within the ODD (Jeong et al., 2002), S-nitrosylation within the ODD (Li et al., 2007b) and SUMOylation in proximity to and within the ODD (Berta et al., 2007).

Additionally, loss of function of different tumor suppressors and gain of function of different oncogenes regulate different steps that lead to HIF activation (Semenza, 2003).

6.1.4. Hypoxia Inducible Factors 2 and 3

HIF1 α is regarded as the principle mediator of the hypoxic response in all mammalian cell types. This hypothesis is underlined e.g. by genome wide expression studies using knock down of HIF1 α. In this study, it was shown that depletion of HIF1 α in fibroblasts was sufficient to prevent induction of hypoxia-dependent genes, while inactivation of HIF-2α was not affecting expression of hypoxic genes (Elvidge et al., 2006). These experiments have been exerted in fibroblasts and the role of HIF2 α is well different in other cell types that are exposed to hypoxia (see below). In fact, HIF2 α shares a high degree of sequence identity with HIF1 α, which is also reflected by their shared ability to heterodimerize with HIF1 β and to bind to HREs to induce transcription of target genes (Raval et al., 2005; Wiesener et al., 2003). Several genes were described to be bound by both isoforms, but often only one of the two is required to activate transcription. This was e.g. confirmed for genes associated with glycolysis (Hu et al., 2007), which were shown to be transcriptionally regulated only by HIF1 α binding. Erythropoietin (EPO) is an example for a gene transcription of which is mainly regulated by HIF2 α (Rankin et al., 2007).

Page 11 Introduction

Moreover, in HIF2 α/VHL loss-of-function studies in mice, using for example a liver- specific HIF2 α knock out model, it was shown, that HIF2 α is an important regulator of hepatic lipid metabolism (Rankin et al., 2009). Additionally, mainly due to its well- described function in cell cycle progression (Gordan et al., 2007), HIF2 α has been implicated to aggressive tumor phenotypes (Qing and Simon, 2009). Generally, the role and importance of HIF2 α, in particular under hypoxic conditions, are less understood compared to HIF1 α and remains to be elucidated. The role and function of HIF3 α, or inhibitory PAS protein (IPAS), which is regulated by several types of alternative splicing, is even less well understood. The most established role for HIF3 α is the ability to form transcriptionally inactive heterodimers with HIF1 α (Makino et al., 2001).

6.1.5. Known transcriptional, HIF1 α-mediated responses

HIF1 α deficiency in mice leads to embryonic lethality at E11. HIF1 α-deficient embryos showed neural tube defects, cardiovascular malformations, and marked cell death within the cephalic mesenchyme (Iyer et al., 1998). In a whole plethora of studies addressing different biological problems in vivo and in vitro over the last decade, HIF1 α was linked to a broad range of genes associated to a variety of physiological processes. Among others, these mainly include basic processes such as angiogenesis, vasodilatation, glucose metabolism, erythropoiesis, oxygen sensing, pH homeostasis, autophagy, development and cell differentiation. A comprehensive list of HIF1 α target genes can be found in a recently published review (Wenger et al., 2005). In the following, two key in vivo functions of Hif1 α in hepatocytes and macrophages are delineated. These functions also build the basis of my study.

6.1.6. HIF1 α in hepatocytes

Oxygen is essential as an electron acceptor in various metabolic functions of the liver. Under normal conditions, oxygen levels of the liver are constantly kept at a high level by a dual blood supply, consisting of the hepatic portal vein and the hepatic arteries and both vessels supply equal amounts of oxygen.

Page 12 Introduction

During the passage through the sinusoids, a periportal-to-perivenous concentration gradient of substrates, products, hormones and oxygen supply is formed due to liver metabolism (Kietzmann et al., 1999). The functional unit of hepatocytes between the aerobic periportal hepatocytes and the anaerobic perivenous hepatocytes, surrounding a hepatic centrilobular vein, spans about 15 to 25 cells (Benhamouche et al., 2006). Oxygen levels drop from 8-9 kPa in the periportal blood to 4-5 kPa in the perivenous blood (Jungermann and Kietzmann, 1996). Expression studies of enriched subpopulations of periportal- and perivenous hepatocytes revealed an increased expression of glycolytic genes in the perivenous hepatocytes, as expected by the oxygen gradient (Braeuning et al., 2006). Additionally, several studies demonstrated the importance of hypoxia in the development and progression of liver diseases. Perivenous hypoxia in particular is regarded to be a major cause for several primary and secondary liver diseases and it is widely believed that perivenous hypoxia can contribute to hepatocellular damage. Perivenous hypoxia plays a crucial role in the etiology of secondary liver diseases such as heart failure (ischemic hepatitis), gut ischemia, indirect drug hepatotoxicity and in the etiology of primary liver diseases such as alcohol-induced liver disease (ALD) or exposure to other xenobiotics like the industrial chemical carbon tetrachloride or the pharmacological agent acetaminophen (Kietzmann et al., 1999). Moreover it was shown, that mice fed with a high fat diet show a deregulation of the hepatic oxygen gradient which gives rise to the progression of NAFLD (non-alcoholic fatty liver disease) to the more serious NASH (non-alcoholic steatohepatitis) (Mantena et al., 2009). Although no study could directly show an increased stabilization of Hif1 α along the metabolic zonation towards the hepatic portal vein, direct functional implications of Hif1 α in liver have recently been revealed by studies in mice with liver specific ectopic expression of a non-degradable form of Hif1 α. The livers of these mice show microvesicular steatosis and a moderately elevated lipid accumulation compared to control livers (Kim et al., 2006). Additionally, a recent study showed an interesting link between Hif1 α-dependent glycolysis and aggressivity of hepatocellular carcinoma (Hamaguchi et al., 2008). Taken together, these studies indicate a crucial function of hypoxia and Hif1 α in normal liver physiology as well as pathophysiology of the liver.

Page 13 Introduction

6.1.7. Hif1 α in macrophages

The energy expenditure of activated macrophages reaches a high level to fulfill its function in immune responses. However, at sites of inflammation, oxygen levels are typically low. Moreover, macrophages experience sustained periods of hypoxia in diseased tissues such as malignant tumors (Vaupel et al., 2001), atherosclerotic plaques (Bjornheden et al., 1999) and arthritic joints (Taylor and Sivakumar, 2005). Therefore, oxygen consumption has to be tightly regulated and macrophages have to rely on anaerobic glycolysis as their major energy source. As described above, Hif1 α is the principal transcription factor to regulate the latter metabolic function. Most importantly, myeloid-specific deletion of Hif1 α revealed that inflammatory processes are impaired due to defects in the glycolytic capacity of macrophages (Cramer et al., 2003). Altered glycolysis resulted in profound impairment of myeloid cell aggregation, motility, invasiveness, and bacterial killing demonstrating the importance of Hif1 α in this cell type. Additionally, under normoxic conditions, Hif1 α can be stabilized by exposure of macrophages to LPS (Jantsch et al., 2008), suggesting a more general importance of Hif1 α in macrophages in host defense against environmental pathogens.

6.1.8. The classical HRE

The canonical DNA motif bound by Hif1 α, consists of a well-conserved 4bp core motif 5’- CGTG-3’. The core motif is part of virtually all described consensus motifs published so far. However, Wenger et al showed that the 5’-CG-3’ of the core motif can be methylated, and therefore can be made inaccessible for Hif1 α (Wenger et al., 2005). Therefore binding studies exploring the affinity of Hif1 α to specific promoters that are based solely on artificially introduced, ‘naked’ DNA such as the luciferase assays, have to be taken with caution. Depending on the study, the core motif can be extended at the 5’ position by an Adenine or a Thymine and a second base in front of the 5’ position of the core motif seems to be preferentially Thymine (Wenger et al., 2005). Apart from this, no extended sequence preference was consistently established so far.

Page 14 Introduction

To circumvent limitations indicated above, I explored Hif1 α promoter binding under native conditions using ChIP-chip or ChIP-Seq (see below). In the following, I aim at introducing general aspects of transcriptional and epigenetic regulation of promoters that are specifically important in the context of latter methods.

6.2. Transcriptional- and epigenetic regulation

The physiological status of every eukaryotic cell is dependent on the regulated production of mRNA by RNA polymerase II (PolII). Transcription is preinitiated by TATA box binding protein (TBP) binding to the promoter. TBP is part of the general transcription factor TFIID, a multimeric protein complex together with thirteen TBP- associated factors (TAFs). The consensus motif bound by TBP is a highly conserved regulatory element called TATA-box with a consensus sequence of TATAA, which is located 28-34bp upstream of the TSS. Recruitment of other general transcription factors and subsequent recruitment of PolII leads to the so called preinitiation complex (PIC). The TATA-box and the Initiator element (Inr), which is defined by the YYANWYY consensus, where the A is at position +1 of the TSS, are the only core promoter elements that, alone, can recruit the PIC and initiate transcription. However, only a low, or basal, rate of transcription is driven by this preinitiation complex and recent genome wide promoter studies revealed that only 10-20% of all mammalian promoters possess a functional TATA-box (Kim et al., 2005). Instead, 72% of all human promoters possess a CpG island (Saxonov et al., 2006), which are stretches in which CG dinucleotides are overrepresented, and it has been shown that only a fraction of CpG-associated promoters have TATA-like elements. Furthermore, CpG-island-associated promoters are most often associated with so-called housekeeping and transcription of these promoters can be initiated over a ~100 bp region resulting in a population of mRNAs that have different lengths but usually the same protein-coding content. Therefore CpG-island- associated promoters are promoters that fall into the ‘broad’ class whereas the promoters that often have TATA and Inr boxes, use only one or a few consecutive nucleotides as TSSs and fall into the `sharp` class (Figure 2) (Carninci et al., 2005).

Page 15 Introduction

In order to further enhance or repress transcription of a specific gene, all required transcription activators have to be present and chromatin structure has to allow for elongation of the transcript by PolII. The precise regulation of these processes is crucial to provide the cell with the required amount of a specific transcript at the right moment. As opposed to 2% of the human genome being protein-coding sequences, one third is believed to be involved in transcriptional regulation, underlining the complexity of this task (for review see (Levine and Tjian, 2003)). Transcriptional regulation can be either indirect by modulation of the chromatin state (see below), or direct by transcription factor interaction with the PIC. Transcription factors and many transcriptional co-activators recognize and bind to a few base pairs spanning, conserved DNA sequences. These motifs are normally located - 6kbp to +4kbp with regard to the TSS. Much of our knowledge about regulatory elements in the PP region is derived by reporter gene assays, which are done by fusing the promoter sequence to a reporter gene and then introducing targeted deletions in that sequence to detect regulatory elements. In order to initiate or modulate transcription, binding sites can also be located distant of TSSs. Indeed, several studies have shown long-range interactions of transcription factors to promoters of regulated genes (Dekker, 2008).

Figure 2: Classification of promoters with respect to the TSS they use . Promoters can be classified in two categories, the sharp type promoter and the broad type promoter. The sharp type promoter often possesses a TATA- box and an Inr element and has a defined TSS. The broad promoter is lacking a classical TATA-Box but often has CpG Islands. The broad type promoter is lacking a defined TSS and transcription is ini tiated from various starting points in front of the coding region. Adopted from (Carninci et al., 2005)

Page 16 Introduction

Apart from these long-range interactions, transcription factors can be found close to clusters of distant localized genes, where transcription is organized in transcriptional factories (Sutherland and Bickmore, 2009). It was also shown that some transcription factors colocalize with sites of active transcription. For example, the progesterone receptor becomes concentrated in nuclear foci only when the hormone ligand is bound, and these foci are associated with active transcription (Arnett-Mansfield et al., 2007). This suggests a physical interaction between clusters of transcription factors and genes, which can be even located on different chromosomes. However, the spatial arrangement of transcription is limited by the fact that large parts of the genomic DNA is bound to the nuclear lamina, organized in so called lamina associated domains (LADs) within the nucleus and active transcription occurs exclusively in non-Lamin associated sites (Guelen et al., 2008). Additionally, as a prerequisite for all binding events, the chromatin status of the binding site must allow the interaction to the DNA binding domain of the transcription factor. According to current knowledge, the chromatin status is mainly regulated by DNA methylation, post-translational histone modifications, chromatin remodeling, histone variant incorporation, and histone eviction (Henikoff et al., 2008; Li et al., 2007).

6.2.1. The role of histones in the regulation of transcription

Histone octamers consist of four types of histones: H3, H4, H2A and H2B. The DNA is wrapped in 1.65 turns of in total 147bp around a histone octamer and linked between octamers by histone H1. Therefore, the most obvious function of histones is the spatial organization of the genetic material within the nucleus. The four subunits can be post- translationally modified in a variety of ways, including phosphorylation, ADP-ribosylation, ubiquitylation, sumoylation, acetylation and methylation. Since many of these modifications are correlated with defined transcriptional responses, the second important function of histones is the regulation of transcription (Kouzarides, 2007; Li et al., 2007; Margueron et al., 2005). .

Page 17 Introduction

Figure 3: Genome -Wide Distribution Pattern of Histone Modifications and phylogenetic tree of JmjC domain containing proteins. (A) Histone modifications and their distribution over the range of a whole arbitrary gene relative to the promoter is shown. The correlation to transcriptional activity is indicated as well as patterns of the histone modification which were determined by genome wide approaches. Adopted from (Li et al., 2007) (B) Phylogenetic tree of all known JmjC domain containing proteins. Putative oncogenes are in red and putative tumor suppressors in green. (JmjC) Jumonji C domain; (JmjN) Jumonji N domain; (PHD) plant homeodomain; (Tdr) Tudor domain; (Arid) AT-rich interacting domain; (Fbox) F-box domain; (C5HC2) C5CHC2 zinc-finger domain; (CXXC) CXXC zinc-finger domain; (TPR) tetratricopeptide domain; (LRR) leucine- rich repeat domain; (TCZ) treble-clef zinc-finger domain; (PLAc) cytoplasmic phospholipase A2 catalytic subunit. Adopted from (Cloos et al., 2008). (C) α-ketoglutarate and iron (Fe) is used as cofactors by JmjC proteins to hydroxylate the methylated histone substrate. To form the highly reactive oxoferryl group which is reacting with the methyl group, Fe(II) has to activate one molecule of oxygen. The spo ntaneous degradation of carbinolamine intermediate leads to the release of one molecule of formaldehyde.. Adopted from (Cloos et al., 2008).

Certain modifications, such as acetylation, are altering the net charge of transcriptional units within the genome, leading to a change from ‘closed’ heterochromatin to an ‘opened’ chromatin state (euchromatin). Of particular interest for transcriptional studies however, is the histone lysine and arginine methylation, since it has been associated to transcriptional activation and repression, heterochromatin-mediated transcriptional silencing, DNA damage response and X chromosome inactivation (Margueron et al., 2005; Martin and Zhang, 2005).

Page 18 Introduction

The best described histone lysine modifications are histone H3 lysines 4, 9, 27, 36 and 79, and histone H4 lysine 20 (Margueron et al., 2005). While trimethylation marks of H3 lysine 4, 36 and 79 are associated with transcriptional activation, trimethylation of H3 lysine 9 and 27 as well as trimethylation of histone H4 lysine 20 is associated with transcriptional inactivation (Berger, 2007). Like all histone modifications, lysine methylations are not unique to one nucleosome. Instead, lysine modifications are spread over the promoter region and often cover the nucleosomes of whole genes (Figure 3A). These patterns are highly dynamic and can be rearranged by histone methylases and histone demethylases. Unlike acetylation, histone methylation and histone demethylation is often catalyzed by a specific enzyme at a specific site resulting in unique functions (Figure 3A).

6.2.2. Histone demethylases

Since N-CH 3 is one of the thermodynamically most stable bonds in nature, the common sense within the epigenetic field was that the only way to revert histone methylation was by histone exchange or by cleavage of the methylated histone tail. With the discovery of the amine oxidase LSD1 as a histone demethylases, this assumption changed (Shi et al., 2004). The LSD1-mediated demethylation process uses flavin adenine dinucleotide (FAD) as a cofactor and can demethylate mono- and dimethylation. In 2006, a second family of histone demethylases, the Jumonji or JmjC domain containing proteins (Tsukada et al., 2006) was discovered. The Jumonji protein family is particularly important in this context as several of its members were shown to be HIF targets (Xia et al., 2009). These proteins contain the conserved JmjC domain, which can demethylate all three methylation states of histones by catalyzing the generation of highly reactive oxygen species (ROS) in the presence of iron, 2-oxoglutarate, and oxygen. The generated ROS attacks the methyl groups on histone lysines and produces unstable intermediate oxidized products that spontaneously release formaldehyde, resulting in the removal of methyl groups from histone lysines (Figure 3C). Of the 27 described proteins with a JmjC domain, 15 possess a known demethylase function of specific lysines or arginines in the H3 tail of histones.

Page 19 Introduction

In a variety of studies, JmjC domain containing proteins were functionally linked to development, differentiation, senescence and X chromosome inactivation. Most recently, histone demethylases have been described to play a crucial role in differentiation and a variety of diseases (Cloos et al., 2008). Therefore it becomes more and more clear that the interplay between histone methylases and demethylases regulate transcriptional responses in a dynamic manner. In fact, our previous view that histone methylation and demethylation constitute stable and irreversible modifications has to be revisited. The first demethylase, which was described to be involved in a dynamic transcriptional response, was JMJD3, a known histone H3 lysine 27 trimethylation (H3K27me3) demethylase. H3K27me3 is associated with transcriptional repression mediated by proteins of the Polycomb group (PcG). JMJD3 can be induced in macrophages upon exposure to bacterial products and inflammatory cytokines mediated by NF κB. The accumulation leads to binding to PcG target genes and regulates their transcriptional activity by removal of the H3K27me3 repressory mark at specific sites (De Santa et al., 2007) providing an intriguing link between inflammation and reprogramming of the epigenome.

6.2.3. Epigenetic modulation under hypoxia

A variety of developmental processes have been linked to hypoxic conditions. For example, it has been speculated that pO 2 in the developing embryo is lower than 2%, implicating an active role of Hif1 α in the embryonic development (Lin et al., 2008). Also the hematopoietic lineage is exposed to hypoxic conditions. By in situ measurements of oxygen levels within the bone marrow of mice, the oxygen levels have been determined to be about 2.4% (Ceradini et al., 2004). Additionally it has been recently shown, that hematopoietic stem cells (HSCs) preferably localize in regions within the bone marrow of low perfusion and vascularization (Parmar et al., 2007). Together, these reports and studies in adipogenic-, myogenic- and chondrogenic differentiation clearly show that hypoxia prevents cellular differentiation and maintains pluripotency of stem/progenitor cells (Lin et al., 2008).

Page 20 Introduction

As of now, the molecular mechanism how hypoxia contributes to these functions is unclear. Since the demethylation reaction mediated by JmjC family members is oxygen dependent (Figure 3C), the general believe is that global methylation is increasing upon hypoxia. However, due to the wide range of affinities to molecular oxygen among different JmjC domain subtypes (Ozer and Bruick, 2007), it is not excluded that certain histone demethylases are active even under severe hypoxia. Only one study addressed the involvement of Hif1 α in the regulation of methylation marks by showing binding to and differential expression of four JmjC domain containing proteins upon exposure to hypoxia (Xia et al., 2009). However two studies were addressing the effect of hypoxia to global methylation levels. Chen et al showed in 2006 that global H3K9me2 levels are enhanced in various cell lines (Chen et al., 2006). More recently Johnson et al analyzed global and gene-specific histone methylation levels (Johnson et al., 2008a). Overall, Johnson et al addressed global levels of four activating methylation marks (H4R3me2, H3K4me2, H3k4me3 and H3K79me2) and 5 repressive marks (H3K27me2, H3K27me3, H3K4me1 and H3K9me2). All modifications tested were 1.4 - 3.6 fold increased upon exposure to 0.2% oxygen for 48 hours on the global level, indicating an enhanced methylation activity or a decreased demethylation activity under hypoxia. However, by a more targeted approach, the promoters of four genes were analyzed for H3k4me3 and H3K27me3 levels upon exposure to hypoxia. Two of these genes were previously shown to be repressed (AFP and Albumin) and two were demonstrated to be enhanced (EGR1 and VEGFA) upon exposure to hypoxia. As expected H3K4me3 levels were enhanced at all promoters 1.6 – 9.2 fold. Surprisingly however, H3K27me3 levels were less than 0.4 times of the levels under normoxia at the promoters of AFP, EGR1 and VEGFA. Although not effecting global methylation levels, this unexpected reduction in methylation levels suggests a possible activity of an H3K27me3 demethylating enzyme at specific sites under hypoxic conditions.

Page 21 Introduction

6.3. Methodology to address Genome wide binding patterns

6.3.1. ChIP-chip

The combination of chromatin immunoprecipitation (ChIP) and DNA microarray hybridization to determine the binding sites of transcription factors in a genome-wide context was first introduced with the study of HNF transcription factors in 2004 (Odom et al., 2004). DNA microarrays consist of unique, single stranded DNA oligonucleotides (features), which are immobilized on a solid surface, in spots with a diameter of a few nanometers. After the ChIP, enriched DNA fragments are linearly amplified by ligation mediated PCR (LM-PCR) and labeled with a fluorescent dye. The labeled DNA fragments are subsequently hybridized to DNA microarrays, which are scanned by a laser to acquire raw intensities of the DNA fragment distribution. The resolution of a CHIP-chip study depends on the fragment size of the DNA, the size of the features and the gap between features on the array. Starting with self-spotted arrays with at most 40000 single features, tremendous progress has been achieved to increase coverage and sensitivity of the arrays. Modern, commercially available DNA- arrays suitable for ChIP-chip studies cover millions of features on one array. The obvious technical limitations introduced by the microarray are mainly probe-specific behavior, dye bias, resolution and design of the array (Johnson et al., 2008b).

6.3.2. ChIP-Seq

To avoid the technical limitation introduced by the DNA-array, high throughput sequencing can be applied to the DNA-fragments derived by a ChIP experiment. The first combination of chromatin immunoprecipitation with genome wide sequencing was established in 2006 (Chen et al., 2008). Since then, several studies successfully applied this method. As opposed to LM-PCR used as an unavoidable amplification step for a ChIP-chip experiment, the amplification step of common ChIP-Seq experiments is far superior in terms of linearity. The specific amplification steps of the ChIPed material is achieved by a manufacturer specific method such as the sequencing-by-synthesis approach of Illumina (Mardis, 2008).

Page 22 Introduction

A genome-wide readout of the protein binding sites is produced by end-sequencing of the amplified and immobilized ChIP fragments. The resulting forward and reverse reads of 36bp (Illumina) are mapped to an existing genome and computationally fused to peaks.

6.3.3. Genome wide binding patterns of transcription factors

Remarkable progress has been made during the past few years in the characterization of transcriptional patterns in a genome-wide scale. The main driving force has been the develop development and improvement of ChIP-chip, ChIP-Seq and other large scale experimental techniques. Therefore, the characteristic binding pattern of several transcription factors, insulators and general transcription factors could be studied on a genome-wide scale including p53 (Chen et al., 2008), PPARg and RXR (Nielsen et al., 2008), estrogen receptor (Carroll et al., 2006), FoxP3 (Zheng et al., 2007), the insulator protein CTCF (Kim et al., 2007), TCF3 (Cole et al., 2008), Polycomb (Pokholok et al., 2005), HNF (Odom et al., 2004), CREB (Zhang et al., 2005), ERRa and ERRg (Dufour et al., 2007) and p63 (Yang et al., 2006). A concise overview over the raw experimental results is depicted in Table 1. Together the binding patterns of different transcription factors are highly varying. Several studies suggested total binding events in the range of hundreds (e.g. Hif1 α), and some suggested total binding to be in the transcription factor to thousands. (e.g. PPARg). The overall overlap of the of nee tod

Table 1: Overview about recent genome wide binding studies

Page 23 Aim of the project

7. Aim of the project

The variety of biological processes, which are affected by a HIF-dependent hypoxic response, highlights the complexity and importance of these transcription factors. Generally speaking, HIF-dependent hypoxic responses mainly entail a shift of energy metabolism towards glycolysis, cell cycle arrest, a decrease in protein translation and induction of neovascularisation factors such as for example VEGF. However, a more global picture of HIF targets and downstream signaling effects is lacking. My work aims at elucidating how eukaryotic cells respond to hypoxia at the molecular level. For this purpose, I generated a global and dynamic regulatory network of the transcription factors HIF-1α in primary hepatocytes and macrophages using ChIP-on-chip in combination with cDNA microarrays and complemented data using a ChIP-Seq approach in a macrophage cell line.

Page 24 Material and Methods

8. Material and Methods

8.1. Media and Buffers used for ChIP and ChIP-chip

Crosslinking Buffer

Stock For 50 ml Final Concentration 1M Hepes-KOH, pH 7.5 2.5 ml 50 mM 5M NaCl 1.0 ml 100 mM 0.5M EDTA, pH 8.0 100.0 l 1 mM 0.5M EGTA, pH 8.0 50.0 l 0.5 mM 37% Formaldehyde 14.9 ml 11% ddH2O 31.5 ml

Block Solution

Stock For 100 ml Final Concentration 10x PBS 10 ml 1X BSA 500 mg 0.5% BSA (w/v) ddH2O 90 ml Total 100 ml Complete Protease Inhibitor Cocktail (Roche) was added

Lysis Buffer 1 (LB1)

Stock For 100 ml Final Concentration 1M Hepes-KOH, pH 7.5 5.0 ml 50 mM 5M NaCl 2.8 ml 140 mM 0.5M EDTA 0.2 ml 1 mM 50% glycerol 20.0 ml 10% 10% NP-40 5.0 ml 0.5% 10% Triton X-100 2.5 ml 0.25% ddH2O 64.5 ml

Page 25 Material and Methods

Lysis Buffer 2 (LB2)

Stock For 100 ml Final Concentration 1M Tris-HCl, pH 8.0 1.0 ml 10 mM 5M NaCl 4.0 ml 200 mM 0.5M EDTA, pH 8.0 0.2 ml 1 mM 0.5M EGTA, pH 8.0 0.1 ml 0.5 mM ddH2O 94.7 ml

Lysis Buffer 3 (LB3)

Stock For 100 ml Final Concentration 1M Tris-HCl, pH 8.0 1.0 ml 10 mM 5M NaCl 2.0 ml 100 mM 0.5M EDTA, pH 8.0 0.2 ml 1 mM 0.5M EGTA, pH 8.0 0.1 ml 0.5 mM 10% Na-Deoxycholate 1.0 ml 0.1% 20% N-lauroylsarcosine 2.5 ml 0.5% ddH2O 93.2 ml

Wash Buffer (RIPA)

Stock For 250 ml Final Concentration 1M Hepes-KOH, pH 7.6 12.5 ml 50 mM 5M LiCl 25.0 ml 500 mM 0.5M EDTA, pH 8.0 0.5 ml 1 mM 10% NP-40 25.0 ml 1% 10% Na-Deoxycholate 17.5 ml 0.7% ddH2O 169.5 ml

Elution Buffer

Stock For 100 ml Final Concentration 1M Tris-HCl, pH 8.0 5.0 ml 50 mM 0.5M EDTA, pH 8.0 2.0 ml 10 mM 10% SDS 10.0 ml 1%

Page 26 Material and Methods

ddH2O 83.0 ml

Linker Oligonucleotides

Oligo JW102 (5’-GCGGTGACCCGGGAGATCTGAATTC-3‘) and Oligo JW103 (5’-GAATTCAGATC-3‘)

Blunting Mix

Stock 1X Mix Final Concentration 10X NE Buffer 2 11.0 L 1x 10 g/ L BSA (NEB) 0.5 L 5 g 10mM each dNTP 1.1 L 100 M 3U/ L T4 DNA polymerase (NEB) 0.5 L 1.5 U ddH2O 41.9 L Total 55 L

Ligase Mix

Stock 1X Mix Final Concentration 5x ligase buffer (Invitrogen) 10.0 l 1x 15 M linkers 6.7 l 2 M 400U/ l T4 DNA ligase (NEB) 0.5 l 200U ddH 2O 7.8 l Total 25.0 l

Mix A

Stock 1X Mix Final Concentration 10X Thermopol buffer (NEB) 4.00 L 1x dNTP mix (2.5 mM each) 5.00 L 250 M oligo JW102 (40 M) 1.25 L 1 M ddH2O 4.75 L Total 15.00 L

Page 27 Material and Methods

Mix B

Stock 1X Mix Final Concentration 10X Thermopol buffer (NEB) 1.0 L 1x Taq polymerase (5U/ L) 0.5 L 0.25 U ddH2O 8.5 L Total 10.0 L

Precipitation Mix

Stock 1X Mix Final Concentration 7.5 M Ammonium acetate 25.0 l 625 mM 100% Ethanol 225.0 l 75% Total 250.0 l

Labeling Mix

Stock 1x Mix Final Concentration 10X dUTP Nucleotide Mix 8.2 L 112/56 nM Cy5- or Cy3-dUTP (1 mM) 1.5 L 17 M Klenow (40 U/ L) 1.5 L 60 U ddH2O 1.8 L Total 13.0 L

8.2. Cell lines

8.2.1. Primary Mouse Hepatocytes

Primary mouse hepatocytes were harvested from male 12-14 weeks old C57BL/6 mice using the protocol of (Seglen, 1976). Collected cells are filtered with a 70µm sieve and washed twice with intermitted spinning steps for 2 minutes at 50 rpm in cold medium. PMH where subsequently plated in DMEM supplemented with 10% FCS (Difco) and 1% penicillin-streptomycin. 2 - 4h after plating the cells are washed once with DMEM.

Page 28 Material and Methods

8.2.2. Primary Peritoneal Mouse Macrophages

Primary Peritoneal Mouse Macrophages were harvested from male 12-14 weeks old C57BL/6 mice. Mice were injected with 2ml of 4% Thioglycollate (Sigma) in the peritoneum and sacrificed after 72h by peritoneal washes with cold PBS. Collected cells are filtered with a 70µm sieve, pelleted and resuspended RPMI supplemented with 10% FCS (Difco) and 1% penicillin-streptomycin. 2 - 4h after plating the cells are washed once with RPMI.

8.2.3. Raw.264 cell line

Raw.264 cells were grown in RPMI (Sigma) supplemented with 10% FCS and 1% penicillin-streptomycin.

8.2.4. Hypoxic conditions

All cells subjected to hypoxia where grown in 15 cm cell culture dishes and 25 ml of growth medium. The Invivo2 400 hypoxia workstation (Ruskinn) was set to 0.5% of oxygen, 37°C and 5% of CO 2. All media used within the hypoxic chamber were preincubated at least for one hour.

8.3. ChIP-on-chip

8.3.1. Preparation of the cells under hypoxia and cross-link proteins to DNA

5 x 10 7 to 1 x 10 8 cells were used for each immunoprecipitation that was used for one ChIP-chip study. On the day of harvesting, cells were incubated in a humid hypoxic chamber (see above) under standard cell culture conditions (37°C, 5% CO 2) for the respective time points. Crosslinking Buffer containing 0.9% formaldehyde was freshly prepared and preincubated under hypoxic conditions for at least 3h before addition to the monolayer. Crosslinking was performed in total for 9 minutes. 1 minute of the crosslinking procedure was carried out under hypoxic conditions and 8 minutes were performed at room temperature and at ambient oxygen levels.

Page 29 Material and Methods

Formaldehyde crosslinking was quenched by the addition of 1/20 volume of 2.5 M glycine to plates. Subsequently cells were rinsed with 5 ml 1X PBS and harvested using a silicone scraper. The cells were then aliquoted into 2 x 50 ml conical tubes and spun at 1,350 x g for 5 minutes at 4°C in a table-top centrifuge with swinging bucket rotor. The supernatant was discarded and cell pellets were flash frozen in liquid nitrogen before storage at -80°C.

8.3.2. Preparation of the magnetic beads

100 l per experiment of Protein G coated Dynal magnetic beads were vigorously resuspended and added to a 1.5ml microfuge tube. 1 ml of Block Solution was added to the tube and beads were gently mixed. Beads were collected using the Dynal small- volume magnetic particle concentrator (Invitrogen) and the supernatant was discarded. After two additional washing steps, 10 g of the respective antibody plus 250 l of ice cold Blocking Solution were added to the beads and incubated overnight at 4°C on a rotating platform. The next day, beads were washed 3x with 1 ml Block Solution as described in above and spun for 1 minute at 4°C at 17,000 x g to collect and remove the supernatant. Finally, beads were resuspended in 100l Block Solution.

8.3.3. Cell lysis

The pellet of approximately 10 8 cells was resuspended in 5ml of Lysis Buffer 1 and rocked at 4°C for 10 min. After spinning at 1,350 x g for 5 minutes at 4°C in a tabletop centrifuge the supernatant was discarded. The pellet was subsequently resuspended in 5ml of Lysis Buffer 2 and rocked gently at room temperature for 10 min. Nuclei were pelleted in tabletop centrifuge by spinning at 1,350 x g for 5 minutes at 4°C and supernatant was discarded. After resuspension of the pelleted nuclei in 3 ml of Lysis Buffer 3 (LB3) cells were transferred to 15ml polypropylene tube that has been cut at the 7 ml mark (to make sonication easier). Sonication of the suspension was performed with a microtip attached to a sonicator (Fisher) at 4°C and on ice, with power settings set to 75% and 30 seconds bursts between 60 seconds of cooling steps in between.

Page 30 Material and Methods

In total the cell suspension was sonicated during 8 of such ON cycles for the normoxic samples and 9 ON cycles for hypoxic cells, to account for increased crosslinking during hypoxia. 300 l of 10% Triton X-100 was then added to the sonicated lysate and mixed by pipetting up and down several times. The lysates were split into two 1.5 ml microfuge tubes and spun at 20,000 x g for 10 minutes at 4°C in a microfuge to pellet debris. Supernatant was then combined from the two 1.5ml microfuge tubes into a new 15 ml conical tube for immunoprecipitation. 50 l of cell lysate was saved from each sample as WCE.

8.3.4. Immunoprecipitation of the chromatin

100 l antibody/magnetic bead mixture were added to 15 ml conical tube containing the cell lysate and were gently mixed overnight on a rotator or rocker at 4°C.

8.3.5. Wash, elution, and reverse cross-linking

One 1.5 ml microfuge tube was pre-chilled for each immunoprecipitate, beads were collected using a magnetic stand and subsequently washed for 7 times with 1ml of Wash Buffer (RIPA). After the last wash beads were washed once with 1 ml TE that contains 50mM NaCl. Cells were spun at 960 x g for 3 minutes at 4°C in a centrifuge and any residual TE buffer was removed a pipette. Elution of the bound Protein DNA complexes was done in 210 l of elution buffer at 65°C for 15min with a Thermoshaker (Eppendorf). During elution, beads were resuspended every 2 minutes by mixing briefly on a vortex mixer. Afterwards beads were spun down at 16,000 x g for 1 minute at room temperature and the supernatant was transferred to a new 1.5 ml microfuge tube. Cross-links of ChIPed eluted fraction and of 50 l of WCE complemented with 3 volumes (150 l) of elution buffer were reversed by incubation in a water bath at 65°C overnight.

Page 31 Material and Methods

8.3.6. Digestion of the cellular protein and RNA

200 l of TE was added to each tube of IP and WCE DNA to dilute SDS in elution buffer. 8l of 10 mg/ml RNaseA (0.2 mg/ml final concentration / Fermentas) were added and mixed and incubated in a circulating water bath for 2 hours at 37°C. 7 l of CaCl 2 stock solution (300 mM CaCl 2 in 10mM Tris pH 8.0) were added to each sample, followed by 4l of 20 mg/ml Proteinase K (0.2mg/ml final concentration / Sigma). The samples were then mixed and incubated in a water bath at 55°C for 30 minutes. 400 l of phenol:chloroform:isoamyl (Fluka) alcohol were added to each tube and samples were thoroughly mixed on a vortex mixer and subsequently centrifuged at 14,000 x g at room temperature for 5 minutes. The supernatant was transferred to a new 1.5ml tube and an equal volume of Chloroform (Fluka) was added. After centrifugation at 14,000 x g for 5 minutes at room temperature the aqueous layer was transferred to a new 1.5ml microfuge tube. A precipitation mix including 16 l of 5M NaCl (200 mM final concentration), 1.5 l of 20 g/ l glycogen (Invitrogen) (30 g total) and 880 l of EtOH were added to each sample. After cooling of the sample at -80°C for at least 30min the mixture was spun at 20,000 x g for 10 minutes at 4°C to create DNA pellets. The pellets were then washed with 500 l of 70% ice-cold EtOH and dried for 10 minutes with a vacuum desiccator. The dried pellets were lysed in 70 l of 10mM Tris-HCl, pH 8.0. While the concentration of the IP samples remains unknown, the concentration of the WCE samples was measured with a Nanodrop (NanoDrop Technologies ) and adjusted to 100ng/ l.

8.3.7. Preparation of linkers for LM-PCR

Oligos JW102 and JW103 were mixed to a final concentration of 40 M each in 250mM Tris-HCl pH 7.9 and 100 l were put into a PCR tube. The linkers were annealed in a Thermal Cycler using the following program:

Step 1: 95 °C 5 minutes

Step 2: 70 °C 1 minutes Step 3: Ramp down to 4°C (0.4°C/min) Step 4: 4°C HOLD

Page 32 Material and Methods

8.3.8. Blunting of the DNA ends and ligation of the linkers

2l (200ng) WCE DNA and 53 l ddH2O were added into a PCR tube. 55 l of each IP sample were transferred into a second PCR tube and on ice. 55 l of blunting mix were added to all samples and cooled for 20 minutes at 12°C in a thermal cycler and subsequently placed on ice. After addition of 11.5 l of cold 3 M sodium acetate and 0.5 l of 20 g/ l glycogen (10 g total) to the sample, Phenol DNA Extraction with subsequent Ethanol precipitation was performed as described above. Pellets were dissolved in 25 l of water. 25 l of ligase mix was added to 25 l of sample and cooled for 16 hours in a thermal cycler set to 16°C. 6l of 3 M sodium acetate and 130 l of 100% EtOH was added to the sample which was then chilled at -80°C for at least 30min. Pelleting the DNA was done by spinning at 20,000 x g for 10 minutes at 4°C. The sample was washed with 500 l of ice-cold 70% EtOH, dried for 10 minutes in a vacuum desiccator and resuspended in

25 l H 2O.

8.3.9. Amplification of the IP and WCE samples

25 l each of IP and WCE DNA were put into separate PCR tubes. 15 l of Mix A was added to each sample and samples were heated in a thermocycler for 2min at 55°C. Then 10 l of Mix B were added to each tube to hot start the reactions with the following PCR program: Step 1: 55°C 2 minutes Step 2: 72°C 3 minutes Step 3: 95°C 2 minutes Step 4: 95°C 30 seconds Step 5: 60°C 30 seconds Step 6: 72°C 1 minute Step 7: GO TO Step 4 x 25 times Step 8: 72°C 5 minutes Step 9: 4°C HOLD After PCR samples were mixed with 250 l of Precipitation Mix each and cooled for 30 minutes at -80°C.

Page 33 Material and Methods

Precipitation was done by spinning at 20,000 x g for 10 minutes at 4°C. Pellets were washed with 500 l of ice-cold 70% EtOH, dried for 10 minutes with a vacuum desiccator, resuspended in 50 l H 2O and concentrations were adjusted to 100ng/ l.

8.3.10. Sample Labeling

Sample labeling and clean up was achieve using Invitrogen’s CGH Labeling kit with a modified labeling procedure: 20.0 l of LM-PCR product (100ng/ L) was put into a PCR tube and 35 l of random primer solution and 20 l of water was added. The sample was mixed on a vortex mixer for 30 seconds, placed in a thermal cycler preheated to 95°C and incubated for 5 minutes. Tubes were then immediately transferred to an ice-water bath and cooled for 5 minutes. Cy5 mix was used for IP DNA and Cy3 for WCE DNA. 3l of the label mix was added in each tube and mixed by pipetting up and down multiple times followed by a 3 hour incubation at 37°C in the dark. The reaction was stopped by adding of 9 l of stop buffer to each tube and subsequent mixing. Samples were transferred to a 1.5 mL microfuge tube and clean up the samples was done using Invitrogen’s CGH column as follows: 0.4ml of Purification Buffer A was added to each tube and mixed with a vortex mixer for 30 seconds. Columns were placed into a 2ml collection tube and spun at 8,000 × g for 1 minute at room temperature. After adding 0.6ml of Purification Buffer B to the column samples were spun in a centrifuge at 8,000 × g for 1 minute at room temperature. Flow- through was discarded and the tube was placed back in the tube. Then 0.2ml of Purification Buffer B was added to the column and the sample was centrifuged at 8,000 × g for 1 minute at room temperature before discarding the flow-through. The purification column was then placed in a new, sterile 1.5-mL collection tube and 50 l of sterile water was added. After incubation at room temperature for 1 minute samples were centrifuged at 8,000 × g for 1 minute at room temperature to elute the labeled DNA.

Page 34 Material and Methods

8.3.11. Sample Hybridization and Scanning of microarrays

All samples were hybridized to Agilent 244k Mouse Promoter Arrays. The hybridization procedure was conducted according to the manufacturer’s recommendations. All slides were treated with the Acetonitrile containing Stabilization and Drying Solution of Agilent after hybridization and before scanning to prevent Ozone degradation of Cyanine 5.

8.3.12. Design of the Agilent 244k Mouse Promoter Arrays

The Agilent 244k Mouse Promoter Arrays consist of 2 slides. Each slide contains 244.000 unique Oligonucleotides with an average length of 60bp and an isothermal design (common melting temperature for all features). The slides represent the Promoters of about 20.000 mouse genes. A promoter is defined as the region of - 5000bp to +2000bp with regard to the TSS. Each gene is covered in average by 25 features.

8.4. ChIP-Seq

ChIP-Seq experiments using the Illumina platform were performed in collaboration by the IGBMC (Strasbourg). To prepare the library, 10 ng of chipped DNA was used (~200 bp DNA fragments linked with 5' and 3' Illumina adapters) using the Illumina kit (Preparing samples for ChIP sequencing of DNA). Library (4 pM) of DNA fragments is then hybridized on the flowcell and clusters are generated using Illumina Cluster Station. Genome Analyzer II (Illumina) is used to sequence (36 cycles).

8.5. mRNA Expression Profiling

TRIzol Reagent (Invitrogen) harvested total mRNA samples were purified using a DNA purification Kit (Machery-Nagel) and subjected to DNAse on column digestion treatment. Total mRNA was quantified using the NanoDrop ND 1000 (NanoDrop Technologies ) and the mRNA integrity was assessed using the Bioanalyzer 2100 (Agilent).

Page 35 Material and Methods

Starting material of total RNA for the amplification, which includes cDNA synthesis using oligodTT7 primer, followed by in vitro transcription, was 1µg. Quality, quantity and dye incorporation control of each cRNA sample was performed using the NanoDrop. Gene expression profiling using the Agilent platform was performed in collaboration by the Functional Genomic Center Zürich.

8.5.1. 1-color array cRNA of the PMM samples were labeled using the Quick Amp Labeling Kit, One-Color (Agilent) and the labeled material was hybridized to Agilent whole mouse genome arrays (G4122F). Arrays were scanned using the Agilent Microarray Scanner G2565BA.

8.5.2. 2-color array cRNA of the PMH samples were labeled using the Quick Amp Labeling Kit, Two-Color (Agilent) and the labeled material was hybridized to Agilent whole mouse genome arrays (G4122F). Arrays were scanned using the Agilent Microarray Scanner G2565BA.

8.5.3. Expression analysis

Raw images of the hybridized arrays were analyzed using the Feature Extraction Software Package Ver. 10.5 (Agilent). Fold changes were calculated using GeneSpring GX Ver. 10 (Agilent). Features with raw intensities lower than 500 were labeled as ´not expressed´ and removed. p-values were estimated using an unpaired students t-test. Features having a p-value > 0.02 were labeled as ´not significant´ and removed. Fold change was calculated using normalized intensities and fold changes lower than 1.5 were labeled as ´not differentially expressed´ and removed.

8.6. Identification of ChIP-chip Peaks

ChIP-chip raw images of the hybridized arrays were analyzed using the Feature Extraction Software Package Ver. 10.5 (Agilent) with the default protocol for ChIP experiments (ChIP_105_Dec08).

Page 36 Material and Methods

Resulting raw intensities were analyzed with ChIP-Analytics Ver.1.3 (Agilent) and inter- arrays median normalization, Intra-array (dye-bias) median normalization and Intra-array Lowess (intensity-dependent) normalization were applied. Peak detection and p-value estimation was performed using the Whitehead Error model v1.0 and the Whitehead Per-Array Neighborhood Model v1.0. Peaks were labeled as ´bound´ if p-values were lower than 0.0006 (for choice of cut off see results).

8.7. Identification of ChIP-Seq Peaks

Peak detection and normalization was done by CisGenome Software v1.1 (Ji et al., 2008). The default parameters for peak detection were used with a window size of 100 bp, cutoff >= 10 reads, step size = 25 bp, maximum gap = 0 and minimum peak length = 0. The detected peaks using these settings were labeled as ´bound´.

8.8. De novo Motif Analysis

8.8.1. Weeder

Weeder (Pavesi et al., 2006) was used to detect overrepresented motifs using the Weeder Cygwin version v1.3.1 with the following settings: ´mouse model´, ´large search´, ´search on both strands´ and ´occurrence can be more than one´.

8.8.2. Gibbs-Motif-Sampler

For low redundant de novo motif search, Gibbs-Motif-Sampler included in CisGenome package was used with the following settings: ´Order of background markov chain´ = 3 and ´No. of MCMC (simulation) iterations´ = 3000.

Page 37 Material and Methods

8.9. Q-PCR Validation of ChIP Hits

RNA was purified from total hearts using TRIzol Reagent (Invitrogen) according to manufacturer’s instructions. 2 g of RNA was used as a template to synthesize cDNA, using Ready-To-Go You-Prime First-Strand Beads (Amersham). Qantitative RT-PCR reactions were set up as recommended by the manufacturer (Roche) and were run an analyzed on the Roche LightCycler 480.

8.10. Annotation of sequences and association of expression- to binding data

All sequences used were annotated with the ENSEMBL release v54 of the 37 NCBI assembly of the mouse genome using the ENSEMBL Core API. All features on the Agilent expression array and all sequences derived by the binding studies were annotated using ENSEMBL Transcript definitions. For comparison of the expression and binding data, ENSEMBL Gene definition was used.

Page 38 Results

9. Results

9.1. Regulation of HIF1 ααα and its target genes

9.1.1. Hif1 α rapidly accumulates upon hypoxia in PMM, PMH

I first aimed at determining ideal time points and conditions for subsequent chromatin- immunoprecipitation of Hif1 α in cells that we wanted to use as a tool in the following ChIP-chip and ChIP-Seq experiments, respectively. Protein abundance of Hif1 α in PMM and PMH was assessed during a time course of hypoxia exposure using Western blotting (Figure 4). In PMH as well as in PMM, Hif1α accumulated most abundantly already after 1.5h and rapidly decreased after 8h.

Figure 4: HIF1 ααα protein is stabilized upon hypoxia in PMM, PMH. Western Blot analysis with 150 µg of nuclear protein extracts from PMM and PMH. Protein levels of HIF1 α were assessed under normoxic conditions and after four time points of exposure to 0.5% oxygen. Lamin A was used as a loading control.

After 16h of hypoxia, Hif1 α was almost undetectable in primary cells. However, exposure of PMH to hypoxia for 8 and 16 hours resulted in reduced abundance of Lamin A (Figure 4) and other reference proteins that I tested (data not shown) most likely due to the fact that the protein translational machinery is inhibited under hypoxia treatment as previously reported (Wouters and Koritzinsky, 2008).

Page 39 Results

9.1.2. Promoter occupancy of Hif1 α target genes correlates well with nuclear protein accumulation

To determine whether nuclear HIF1 α accumulation correlates with its promoter occupancy, I analyzed three established HIF1 α-targeted promoters at different time points of hypoxia by ChIP-QPCR. PMM and PMH showed a significant increase of HIF1 α promoter binding after 3h of hypoxic exposure (Figure 5A-B).

Figure 5: Expression of HIF1 ααα target genes was increased several hours after the initial binding event. A- B: ChIP-QPCR of three established HIF1 α binding locations at the promoters of GAPDH, LDHA and JMJD1A were performed. ChIP was performed from PMM (A) and PML (B) after different time points of exposure to hypoxic conditions (0.5% O2). Ct values were normalized to percent of input and afterwards to normoxia. C - D: RT-PCR was performed of total mRN A from PMM (C) and PML (D) after different time points of exposure to hypoxic conditions (0.5% O 2).

Page 40 Results

In both cell types and for all three tested genes, promoter occupancy was maintained after 16h of hypoxic exposure at levels seen upon 3h of hypoxic conditioning indicating that HIF1 α stabilization leads to its binding to promoters in PMM and PMH.

9.1.3. Expression of the aforementioned HIF1 α target genes was enhanced several hours after the initial binding event

To assess whether promoter occupancy correlates with changes in expression of respective genes, transcript levels of genes were measured for PMM (Figure 5C) and PMH (Figure 5D) at different time points after hypoxic exposure using quantitative RT- PCR. As expected, induction of transcripts was observed several hours after HIF1 α promoter occupancy for all three tested transcripts and in both tested cells. The expression dataset with highest resolution revealed that a plateau phase is reached approximately after 8h-16h of exposure to hypoxia (PMH in Figure 5D). These experiments demonstrated that expected hypoxic responses occur in PMH and PMM under conditions used making it to a suitable system for subsequent experiments. Based on the pattern of promoter occupancy and subsequent transcription, we decided to take the 3h time point for chromatin-IPs and 16 h for the expression analysis (Figure 6).

Figure 6: Experimental design based on promoter occupancy and expression patterns. PMH and PMM will be exposed to 0.5% and 21% Ox ygen levels. ChIP and whole mRNA extraction will be performed for each condition individually. Hypoxic conditions will be applied for 3h and 16h for the binding study and the expression study, respectively.

Page 41 Results

9.2. Genome wide expression and binding studies

9.2.1. ChIP-chip

To assess the binding events of Hif1 α on a genome-wide level, I performed ChIP-on- chip assays with extracts of PMH and PMM. For hybridization, we chose commercially available mouse promoter arrays (Agilent) that allowed for assessment of binding events of Hif1 α across promoters (-5kbp to +2kbp with regard to the TSS) of the entire genome (Figure 6).

9.2.2. HIF1 α promoter occupation

To determine the amount of significant binding events, I applied stringent statistical criteria with a p-value of <0.0006, as determined by ChIP-QPCR (see chapter 9.2.3). Comparison of the individual hypoxia- to normoxia control experiments revealed that the amount of significantly bound genes was less than 25% of the binding events observed under hypoxia. In general the statistical significant peaks measured in cells under normoxic showed less bias towards the TSS, less enrichment for HRE and a approximately ten fold lower raw intensity level compared to their relative hypoxic datasets (data not shown) suggesting nonspecific enrichment. However, around 16% of the binding events observed in hypoxia experiments were found in cells under normoxic conditions only, suggesting nonspecific enrichment for most of the genes.

Page 42 Results

Figure 7: HIF1 ααα targets distinct genes in different cell types. (A) Significant binding events of HIF1 α can be divided into three subclasses. Group I peaks are peaks that are uniquely found and that can be associated to a unique position within the genome. Group II peaks are pe aks that can be associated to a maximum of two genes (one upstream and one downstream within either the whole genome or the PP region) or minimal to one gene. All peaks that cannot be associated to a unique gene within the respective region are neglected f rom Group II. On the other hand a peak can be found twice if a peak is associated to a gene up- and downstream (e.g. at bidirectional promoters). Since Agilent expression arrays cover only a limited amount of ENSEMBL annotated transcripts, Group II peaks c an be subdivided to a group of peaks that are covered by the Agilent expression arrays and have therefore present expression data (Group III). The overlap between each dataset is represented by Venn Diagrams on the right.

The amount of all statistically significant binding events under hypoxic conditions of each experimental set are indicated in group I (Figure 7). The number of genes in group I that could be associated to current ENSEMBL genes within the proximal promoter (PP), are indicated in group II. Group III genes are targets in group II that are also represented on arrays used for my genome-wide expression arrays. (Figure 7). A complete list of group III genes is provided in the appendix.

Page 43 Results

9.2.3. Validation of ChIP-chip experiments

In order to validate ChIP-chip data, ChIP-QPCR of a randomized set of significantly bound target genes in PMH (Figure 8A) and PMM (Figure 8A and B, respectively) was performed. In total, one gene was randomly chosen out of one group of 50 genes, ranked according to decreasing ChIP-chip data scores (p-value) until rank number 500. Additionally, three well established targets as described by Wenger et al (Wenger et al., 2005) and four negative controls were validated. ChIP samples of each IgG control, normoxia and hypoxia were included. The enrichment levels confirmed the ChIP-chip data approximately until group number 5 (top 250 genes) for both datasets in PMH and PMM. Enrichment levels of groups lower than group number 5 were in general indistinguishable (pvalue > 0.1) compared to the mean enrichment scores of the negative control genes as well as the IgG ChIP. To compare datasets in a statistically defined manner, ChIP-chip p-Values were lowered until the first 5 groups were covered within both datasets. The adjusted ChIP-chip p- Value used for all ChIP-chip experiments was < 0.0006. In total, 3 of 16 and 1 of 14 tested peaks within the first 5 groups of PMH and PMM dataset, respectively could not be confirmed by ChIP-QPCR. Therefore, a false discovery rate (FDR) of 20% and 14% for PMH and PMM respectively can be estimated. 6 out of 16 and 2 out of 14 tested genes of the first 5 groups in PMH and PMM respectively, were also significantly bound under normoxia (data not shown). Two genes in the PMH under normoxia but none in the PMM dataset could be confirmed by ChIP- QPCR.

Page 44 Results

Figure 8: ChIP -chip data validation by ChIP -QPCR. ChIP has been performed with PMH (A) and PMM (B). Primers were chosen within the region of enrichment of indicated genes. Results were expressed as percentage of input. Four control regions were included as negativ e controls. The first bar of each gene represents an IgG control ChIP, the second bar HIF1 α ChIP with cells under normoxic conditions and the third bar HIF1 α ChIP with 0.5% hypoxia treated cells. The level of significance of the comparison between negative control targets and ChIP-chip results is indicated by an asterisk. P-values were calculated using an unpaired, two-tailed students t - test. The level of average hypoxic enrichment levels in negative control HIF1 α ChIP is indicated by the red line.

Page 45 Results

9.2.4. Hif1 α binds to a distinct subset of genes in PMM and PMH

The binding data of PMH and PMM overlaps by about 20% (46 genes) between the two cell types (Figure 9) indicating a strong cell type specificity of Hif1 α. I next compared genes, that show modulated expression in PMH and PMM under hypoxic conditions

Figure 9: Overlap between PMH and PMM binding data. All group III gene sets of the primary cell ChIP-chip experiments were represented with Venn diagrams and compared.

9.2.5. Expression levels differ in PMM and PMH exposed to 16h of hypoxia

Genome-wide expression levels of PMM and PMH in response to 16h of exposure to hypoxia were measured. Commercially available mouse whole-genome expression arrays (Agilent) were used. A transcript was termed ´up regulated or down regulated´ if raw intensity values exceeded a mean value of 500 and significant if the p-value was lower than 0.02. Expression levels of at least ten randomly chosen transcripts were validated by quantitative RT-PCR (data not shown). The estimated FDR of the PMM and the PMH dataset is 10%. In total 377 and 1217 genes were more than 1.5 fold up regulated in PMH and PMM, respectively (Figure 10). The list of the top 300 up regulated genes is provided in the appendix.

Page 46 Results

Figure 10 : Overlap between expression data o f PMH and PMM. Venn diagrams represent the amount of differentially expressed genes in PMH and PMM as indicated, after 16h of exposure to hypoxic conditions.

9.2.6. Bound genes marginally overlap with differentially expressed genes in PMM and PMH

To determine the frequency, at which Hif1 α binding resulted in differential expression, the expression and the binding data sets of PMH and PMM were compared. Of the 176 significantly bound genes in the PMM dataset, only 44 genes were enhanced upon hypoxia (Figure 11A). Within the PMH dataset, only 48 genes were induced upon hypoxia. Reduced expression upon hypoxia was observed only for one gene, within the PMM dataset and for a total of 11 genes within the PMH dataset (Figure 11B). These data suggest that Hif1 α binds a large portion of genes that are not differentially expressed upon hypoxia and most genes that show altered expression upon hypoxia are not bound by HIF As already presented, the binding data of PMH and PMM overlaps by about 20% (46 genes) between the two cell types (Figure 9) indicating a significant overlap (24.6 fold enrichment compared to a random gene set) but also a strong cell type specificity of Hif1 α. I next compared genes, that show modulated expression and are bound by Hif1 α in PMH and PMM under hypoxic conditions. I found a 43% (20 genes) overlap of genes that were differentially expressed and that were directly bound by Hif1 α in both cell types (Figure 11C). A complete list of these genes including expression data and significance of the binding event is presented in (Table 2).

Page 47 Results

Taken together, these two Hif1 α location studies show a clear enrichment in common targets in different cell types. However, the majority of Hif1 α target genes within each data set do not overlap, suggesting a strong cell-type specific promoter binding of Hif1 α.

Figure 11 : Overlap between expression data and binding data of PMH and P MM. Green (up regulated) and red (down regulated) Venn diagrams represent the amount of differentially expressed genes in PMM (A) and PMH (B) as indicated, after 16h of exposure to hypoxic conditions. Blue Venn diagrams indicate group III binding data. (C) Comparison of the intersection of (A) and (B)

Page 48 Results

Table 2: Genes bound by HIF1a and induced by hypoxia in both, PMM and PMH. The genes listed in this table are satisfying statistical criteria for significant binding and transcription al up regulation in both, PMM and PMH. Fold change levels are color coded with low fold change (yellow) to high fold change (red). Down regulated genes are color coded in blue.

9.2.7. Functional clustering reveals common and specific biological roles for Hif1 α in PMM and PMH

To classify biological processes associated with enriched genes, the genes of group II of both PMH and PMM cells were clustered into overrepresented Gene Ontological Clusters of Biological Processes using DAVID Bioinformatics Resource (http://david.abcc.ncifcrf.gov/) (Table 3). The principal category, which was significantly overrepresented in all tested cell types, was glycolysis (highest p-value < 0.0013). Additionally, vasculature development could be associated to PMH only, whereas transcription and regulation of transcription could be associated to PMM. These findings confirm that Hif1 α is essential in the regulation of glycolysis under hypoxic conditions in hepatocytes as well as macrophages. This highlights the importance of anaerobic glycolysis for both cell types. However, these results also imply cell type-specific regulation of biological processes by Hif1 α. Angiogenesis is a more important response to hypoxia in PMH compared to PMM. Instead, PMM regulate a high number of genes associated to regulation of transcription.

Page 49 Results

Table 3: The majority of HIF1 ααα targets can be associat ed to glycolysis using DAVID and GO categories. The complete group II genes of PMH and PMM were analyzed by DAVID and clustered into GO categories of biological processes (category BP4). The top 10 overrepresented clusters are presented along with the resp ective count of Genes, pvalue, fold enrichment within the cluster and False Discovery Rate as computed by DAVID.

9.2.8. HIF1 α binding occurs preferably in close proximity of the TSS

To check for preferred binding sites of HIF1 α within promoter regions, all group II sequences were summed up in groups of 100bp and plotted according to their distance to the TSS within the range from -5000bp to +3000bp. Indeed, a strong bias of the binding events towards the TSS could be observed in PMM (Figure 12A) and PMH (Figure 12B) and no strong bias could be observed in IgG control ChIP samples (Figure 12C). Additionally, the binding events occurred preferably before the TSS in all cell types. Within the PP region, no binding site within enhancer regions was allocated. Together these results demonstrate that the binding pattern of Hif1 α in all tested cell types follows the typical binding pattern of a transcription factor with a strong bias towards the TSS (Xia et al., 2009). A preference for enhancers at specific sites could not be observed. Additionally, a higher number of simultaneous binding events could be observed among genes with higher binding scores.

Page 50 Results

Figure 12 : HIF1a binding occurs preferentially in close proximity to the TSS. Group II peaks of PMM (A), PMH (B), and IgG control (C) were clustered into groups of 100bp and plotted according to their relative distance to the TSS and against the frequency of total genes within the whole PP region.

Page 51 Results

9.2.9. Binding peaks of differentially expressed genes show a bias towards the TSS

In order to test whether the proportion of Hif1 α binding events that caused alteration of transcript levels shows a bias towards a preferred location on the respective promoter region, the frequency of bound and at the same time differential expressed genes was plotted against their relative location on the promoter. Therefore, I clustered the binding events into groups of 1kpb or 500bp and within each group, the frequency of bound genes with differential expression was calculated and all groups with less than 15 genes were neglected (Figure 13). In PMH and PMM, it was evident that the amount of bound genes with differential expression is almost twice as high within the range of 1000bp around the TSS, compared to groups that were located elsewhere. Furthermore, within the IgG control ChIP-chip dataset, no such bias could be observed. This analyses demonstrate, that genes proximally bound by Hif1 α with regard to the TSS, are more likely differentially expressed as compared to the distally bound genes

Page 52 Results

Figure 13 : Peaks of differentially expressed gen es show a bias towards the TSS. Genes of the PMM and IgG group III datasets were clustered into groups of 1000bp and genes of the PMH dataset were clustered into groups of 500bp. The groups were plotted according their relative location to the TSS and agai nst the frequency of total genes within the whole PP region that showed differential expression.

Page 53 Results

9.3. Genome wide binding study using the murine leukemic monocyte- macrophage cell line (Raw.264) and ChIP-Seq

Hif1 α has been demonstrated to regulate a vast majority of genes that are induced by hypoxia (ref). The relatively small amount of significant binding events in the ChIP-chip experiments for primary cells as opposed to changes in expression of many more genes may indicate that the latter technique and tools attached to it do not quite provide a full picture about HIF binding. In any case, the number of genes revealed by ChIP-chip does not contain sufficient sequence information to perform valuable in silico analysis of promoters that are bound by Hif1 α. In order to complement the promoter-biased binding data of PMM with genome-wide data and to validate the results with a different method, ChIP-Seq was performed with Raw.264 cells (Figure 14). As explained above, this assay is not dependent on the resolution of the promoter arrays.

Figure 14 : Experimental design based on promoter occupancy and expression patterns. Raw.264 cells will be exposed to 0.5% and 21% Oxy gen levels. ChIP will be performed for each condition individually with hypoxic conditions of 3h. Resulting ChIP fragments will be sequenced and reads will be mapped to reference genome.

Page 54 Results

9.3.1. Hif1 α rapidly accumulates upon hypoxia in Raw.264 cells

As described for primary cells, I first aimed at determining ideal time points and conditions for subsequent chromatin-immunoprecipitation of Hif1 α in Raw.264 cells. Protein abundance of Hif1 α in Raw.264 cells was assessed during a time course of hypoxia exposure using Western blotting. Nuclear Hif1 α in Raw.264 cells was detectable at 1.5h after exposure to hypoxia and gradually increased during following indicated time points (Figure 15).

Figure 15: HIF1 ααα protein is stabilized upon hypoxia in Raw.264 cells. Western Blot analysis with 150 µg of nuclear protein extracts from Raw.264 cells. Protein levels of HIF1 α were assessed under normoxic conditions and after four time points of exposure to 0.5% oxygen. Lamin A was used as a loading control.

To determine whether nuclear HIF1 α accumulation correlates with its promoter occupancy, I analyzed three established HIF1 α-targeted promoters at different time points of hypoxia by ChIP-QPCR. All three tested genes showed a significant increase of HIF1 α promoter binding after 3h of hypoxic exposure (Figure 16A). For all three tested genes, promoter occupancy was maintained after 16h of hypoxic exposure at levels seen upon 3h of hypoxic conditioning. These results indicate that HIF1 α stabilization leads to its binding to promoters in Raw.264 cells.

Page 55 Results

Figure 16: The Expression of HIF1 ααα target genes was induced several hours after the initial binding event. A: ChIP-QPCR from Raw.264 cells of three established HIF1 α binding locations at the p romoters of GAPDH, LDHA and JMJD1A and at different time points of exposure to hypoxic conditions (0.5% O2) were performed. Ct values were normalized to percent of input and afterwards to normoxia. B: RT-PCR was performed of total mRNA from Raw.264 cells after different time points of exposure to hypoxic conditions (0.5% O 2).

9.3.2. Expression of the aforementioned HIF1 α target genes was enhanced several hours after the initial binding event

To assess whether promoter occupancy correlates with changes in expression of respective genes, transcript levels of genes were measured for Raw.264 cells (Figure 16B), at different time points after hypoxic exposure using quantitative RT-PCR. As expected, induction of transcripts was observed several hours after HIF1 α promoter occupancy for all three tested transcripts and in all tested cells. These experiments clearly demonstrated that expected hypoxic responses occurs in Raw.264 cells under conditions used making it to a suitable system for subsequent experiments. As for the primary cells tested earlier, based on the pattern of promoter occupancy and subsequent transcription, we decided to take the 3h time point for chromatin-IPs.

Page 56 Results

9.4. Promoters that are occupied by HIF1 α in Raw.264 cells

9.4.1. Cell-specific binding of targets by Hif1 α under hypoxic conditions

To determine the amount of significant binding events, I applied stringent statistical criteria to the ChIP-Seq data set. Comparison of the individual hypoxia- to normoxia control experiments revealed that within all three normoxia data sets, the amount of significantly bound genes was less than 25% of the binding events observed under hypoxia. Additionally, around 16% of these binding events were found in cells under normoxic conditions suggesting nonspecific enrichment for most of the genes.

Figure 17: HIF1 ααα targets distinct genes in different cell types. Significant binding events of HIF1 α can be divided into three subclasses. Group I peaks are peaks that are uniquely found and that can be associated to a unique position within the genome. Group II peaks ar e peaks that can be associated to a maximum of two genes (one upstream and one downstream within either the whole genome or the PP region) or minimal to one gene. All peaks that cannot be associated to a unique gene within the respective region are neglect ed from Group II. On the other hand a peak can be found twice if a peak is associated to a gene up- and downstream (e.g. at bicistronic promoters). Since Agilent expression arrays cover only a limited amount of ENSEMBL annotated transcripts, Group II peaks can be subdivided to a group of peaks that have are covered by the Agilent expression arrays and have therefore present expression data (Group III). The overlap between each dataset is represented by Venn Diagrams on the right.

Page 57 Results

The amount of all statistically significant binding events under hypoxic conditions of each experimental set are indicated in groups as described for ChIP-chip studies described above (Figure 17). Group II and III genes are subdivided into A and B, where A stands for binding events associated to genes within an extended promoter region (-100kbp to +100kbp) and B for binding events associated to genes that are located within the limits of the PP region. In total 3320 genes of in total 27078 genes within the database, could be associated to 8245 peaks of the Raw.264 cell dataset within the PP region. A complete list of the first 300 group III genes is provided in the appendix.

9.4.2. ChIP-Seq data validation

In order to determine the significance of the Raw-264 cell ChIP-Seq experiment considering all 8245 binding events referred to as group I, I calculated mean conservation of the area under the peak and compared the data to location- and size matched, randomized control region (LSC) derived sequences (Figure 18A). Sequences were ranked according to their binding scores (amount of reads) and the mean conservation score was calculated for groups of 250 sequences. Throughout the dataset and until the last group, the mean conservation score was significantly elevated compared to the one of the LSC sequences (pvalue<0.001). Another data consistency assessment was performed by analysis of the peak sequences. Sequences of the ChIP-Seq experiment were ranked according to their binding scores (amount of reads) and the occurrence of the classical HRE within a group of 50 sequences was traced (Figure 18B). Finally, enrichment levels for each motif were calculated compared to a LSC sequence. Again, significant (pvalue<0.001) and more than 10 fold enrichment was observed for HREs considering the whole ChIP-Seq dataset, whereas no enrichment was observed for the repeat like motifs which were used as controls. Taken together, conservation- and motif analysis revealed that 8245 peaks were statistically significant (pvalue<0.001).

Page 58 Results

Figure 18: ChIP-Seq data validation. (A) Raw.264 HIF1 α ChIP-Seq data were ranked according to their score and clustered into groups of 300 peak. Average PhasCons score was plotted against the ranked groups. The green line and the blue line represent the ChIP-Seq data and LSC sequences respectively. The black line represents a sixth order polynomial regression curve through each dataset. The dashed red line represents a widely accepted cut off for significantly enhanced PhasCons conservation scores. (B) Raw.264 HIF1 α ChIP-Seq data were ranked according to thei r score and clustered into groups of 50 peaks. Four highly overrepresented motifs within the ChIP-Seq dataset of Raw.264 cells were computed by Gibbs-Motif-Sampler. The average fold overrepresentation compared to an LSC sequence set was computed and plotte d against the ranked groups. The blue line represents the overrepresentation of an HRE like motif, which was one of the top motifs given by Gibbs - motif-Sampler. The black line represents a sixth order polynomial regression curve through each dataset.

Page 59 Results

9.4.3. Functional clustering of the ChIP-Seq data

To characterize biological processes associated with enriched genes, the top 500 genes of group IIB were clustered into overrepresented Gene Ontological Clusters of Biological Processes using DAVID Bioinformatics Resource (http://david.abcc.ncifcrf.gov/) (Table 4). The principal category, which was significantly overrepresented in all tested cell types, was glycolysis. These findings confirm that Hif1 α is essential in the regulation of glycolysis under hypoxic conditions in Raw.264 cells. Additionally, a high number of genes clustered to regulation of transcription. Comparisons of single genes and associated functions are provided below.

Table 4: The majority of HIF1 ααα targets in Raw.264 cells can be associated to glycolysis using DAVID and GO categories. The top 500 genes derived by the ChIP-Seq data of Raw.264 cells were analyzed by DAVID and clustered into GO categori es of biological processes (category BP4). The top 10 overrepresented clusters are presented along with the respective count of Genes, pvalue, fold enrichment within the cluster and False Discovery Rate as computed by DAVID.

9.4.4. HIF1 α binding occurs preferably in close proximity of the TSS

To check for the preferred binding sites of HIF1 α within the genome, all group II genes were summed up in groups of 100bp and plotted according to their distance to the TSS within the range from -5000bp to +3000bp (Figure 19A). Raw.264 cells showed a strong bias towards the TSS, with 35% of all binding events being located within a -100bp to +100bp window. This bias was even more pronounced at an extended window from - 100kbp to +100kbp in groups of 1kbp (Figure 19B). Within the PP region, no striking enhancer like, preferred binding site could be located. The genome wide peak distribution of HIF1 α in Raw.264 cells revealed, that a significant proportion of binding events occurs within the CDS (Figure 20). The distribution between intergenic and intragenic associated peaks was not significantly different compared to a randomized dataset.

Page 60 Results

Figure 19: HIF1 ααα binding occurs preferentially in close proximity to the TSS in Raw.264 cells. (A) Group IIA peaks and Group IIB peaks (B) of Raw.264 cells were clustered into groups of 100bp and plotted according to their relative distance to the TSS and against the frequency of total genes within the whole PP region.

Apart from the bias towards the TSS as described above, a two times overrepresentation compared to a randomized dataset of simultaneous binding to the 5’ and the 3’ end of genes was observed (Figure 20). Although, the 314 genes showing this pattern among the group II genes of the Raw.264 dataset did not show any bias in expression levels, the mean binding score was clearly higher. The genome wide peak distribution of HIF1 α in Raw.264 cells revealed, that a significant proportion of binding events occurs within the CDS (Figure 21). The distribution between intergenic and intragenic associated peaks was not significantly different compared to a randomized dataset. Together these results demonstrate that the binding pattern of Hif1 α in Raw.264 cells follows the typical binding pattern of a transcription factor with a strong bias towards the TSS. A preference for enhancers at specific sites could not be observed. Additionally, among the genes with higher binding scores, a higher number of simultaneous binding events could be observed.

Page 61 Results

Figure 20: Example of the overrepresented pattern showing simultaneous HIF1 ααα binding to the 5’ and the 3’ end. The ChIP-Seq results of the gene TFRC within the Raw.264 cell dataset were plotted according to its genomic location. The black line as well as the heat plot represents the Hypoxia ChIP-Seq reads and the red line represents the Normoxia HIF1 α ChIP-Seq experiment. Conservation scores as well as genomic region are shown below.

Figure 21: Genome wide HIF1 ααα binding distribution relative to functional genomics elements in Raw.264 cells. Group I peaks were annotated and associated to functional genomics elements. The frequency of the overall amount of pe aks was calculated and the significantly overrepresented elements compared to the general genomic element distribution within the mouse genome, is marked in red.

Page 62 Results

9.5. Characterization of the Hypoxia Response Element

9.5.1. Hif1 α binds preferably to the extended core HRE -CGTACGTGC- motif.

To characterize motifs that are targeted by HIF1 α, Weeder, an algorithm, that was shown to predict overrepresented motifs in mammalian genome-wide association studies with high accuracy (Linhart et al., 2008), was used. In all tested cell types and for both ChIP-chip and ChIP-Seq the canonical HRE -ACGTG- was found as the most overrepresented motif (Figure 23A). Importantly, the published consensus motif could be refined. At position -3 to -1 to the core HRE a -CGT- was clearly overrepresented in all three experiments. Based on ChIP-Seq results, at position +1 of the core HRE, an additional -C- appears to defiine the HIF1 α binding site. Conclusively, the published core HRE -CGTG- was confirmed and could be expanded to a more precise -CGTACGTGC-.

9.5.2. HRE harboring Peaks are preferably localized close to the TSS

In order to visualize localization of the HRE within the promoter region in relation to binding sites, I plotted the promoter region against the location-sorted peaks. HREs were highlighted show that the HRE positive sequences display a clear bias towards the TSS across all tested cell types (Figure 22). This observation was especially true for PMM; and Raw.264 cells showed an even more pronounced TSS-biased distribution of HRE harboring peaks. Analysis of IgG control ChIP data revealed a randomized distribution of peaks and HRE motifs relative to the peaks. Altogether, these data show that HREs occur less frequently in peaks distant to the TSS (>2kbp6). Since I showed above that the overall peak distribution is strongly TSS-biased in all tested cell types (Figure 12 and Figure 19), peaks without a clear HRE and more distant localization to the TSS are most likely non-specific. In order to visualize localization of the HRE within the promoter region in relation to binding sites, I plotted the promoter region against the location-sorted peaks. HREs were highlighted show that the HRE positive sequences display a clear bias towards the

Page 63 Results

Figure 22 : Peaks harboring HRE elements are preferably localized clo se to the TSS. In order to visualize the general HRE and peak distribution, each peak of each dataset was plotted sorted according to the distance to the TSS against the genomic location. The genomic region is represented by a black line, whereas a peak region is represented by a green line. A HRE consensus site is represented as a blue spot and red if it is located within a peak region. Once a peak is harboring a HRE, the green line is represented in yellow instead. (A): Raw.264, (B): PMM, (C): PMH, (D): IgG

Page 64 Results

Figure 23 : The most overrepresented motifs detected by Weeder and Gibbs -Motif -Sampler are closely related to the established HRE. All datasets were tested for overrepresented motifs by Weeder (A) and Gibbs - Motif-Sampler (B). In the first Row, the well-established HRE is shown (Wenger et al., 2005). In the following rows the top motifs of each analysis are represented with a Weblogo Plot expressed in bits of likelihood. Additionally, for the motifs detected by Weeder, the pvalue of the probability comparison to a matched, randomized region is given.

Page 65 Results

9.5.3. Tandem HREs are commonly found across all tested cell types

Since Gibbs Motif Sampler allows for larger candidate motif searches compared to Weeder, it was used as an additional tool to detect overrepresented motifs. Additionally, it allows for better identification of flanking sequences that are potentially occupied by other transcription factors (Figure 23B). The resulting motifs of this algorithm are much less redundant compared to the ones resulting from the Weeder analyses. However, to assess the significance of each motif, the input sequences have to be compared to a reference genome. The references used in this study were LSC sequences. Within the PMH and PMM dataset, the core HRE frequently occurred as a duplet: - CGTGNNNNACGTG-. Analysis of the Raw.264 sequences resulted in a smaller, single HRE similar to the motif resulting from the analysis by Weeder. One reason that Gibbs Motif Sampler failed to detect tandem HREs as one of the most common Hif1 α target motifs within the Raw.264 dataset, can be the in average 10 times smaller peak size. To test this hypothesis, I analyzed broadened ChIP-Seq peaks (+50bp) of the Raw.264 dataset with Gibbs Motif sampler for overrepresented motifs. Indeed, one of the most overrepresented consensus sites compared to LSC sequences was the tandem HRE consensus site -CGTGNNNNNCGTG- (Figure 23B). As the single HREs, the tandem HRE harboring peaks, which are around 15% of all peaks in all datasets, show a bias towards the TSS (Figure 24B). However, due to the 10 times smaller peak size, within the Raw.264 dataset the tandem HRE frequency per base pair is far higher than within the primary cell datasets. Conclusively, these data suggest that tandem HREs are frequently observed in all tested cell types and broaden the view on the sequence preference of Hif1 α-mediated hypoxia regulated genes.

Page 66 Results

Figure 24 : Tandem HREs increase ChIP efficiency and are commonly found across all tested cell types . (A) Raw.264 HIF1 α ChIP-Seq data were ranked according to their score and clustered into 50 peak bins. Three highly overrepresented motifs within the ChIP-Seq dataset of Raw.264 cells were computed by Gibbs-Motif - Sampler. The average fold overrepresentat ion compared to an LSC sequence set was computed and plotted against the ranked bins. The green line represents the overrepresentation of an tandem HRE motif, which was one of the top motifs given by Gibbs-motif-Sampler. The black line represents a sixth o rder polynomial regression curve through each dataset. (B) In order to visualize the tandem HRE distribution, each peak of each dataset was plotted and sorted according to the distance to the TSS against the genomic location. Due to the bigger size of the tandem HRE, peak regions of the ChIP-Seq region were broadened by 50bp to each side. The genomic region is represented by a black line, whereas a peak region is represented by a green line. A tandem HRE site is represented as a blue spot and red if it is l ocated within a peak region. Once a peak is harboring a tandem HRE, the green line is represented in yellow instead.

Page 67 Results

9.6. Transcription factors interacting with Hif1 ααα

9.6.1. AP1 transcription factor motifs are overrepresented at enhancer regions bound by Hif1 α in Raw.264 cells

As mentioned above, motif search results by Gibbs Motif Sampler offer a variety of non- redundant candidate consensus sites. To detect overrepresented motifs, I checked several of those candidate consensus sites within the datasets and compared them to LSC sequences. One motif that showed a consistent overrepresentation throughout the whole dataset of Raw.264 cells was the canonical AP1 transcription factor target site - TGANTCA- (Figure 25A). A comparative blot with other candidate motifs showed a significant enrichment throughout the ranked and clustered peaks (Figure 25A). Interestingly, the highest enrichment scores for the AP1 consensus are found within groups larger then rank 1000. A location plot of a 788 peaks window underlined this observation (Figure 25B). Unlike the HRE positive peaks, the Hif1 α-bound peaks that harbor an AP1 side are located far away from the TSS.

9.6.2. Potential interplay of HIF1 α and AP1 may underlie developmental and differentiation processes

In order to assess functional aspects of AP1 and Hif1 α co-regulation, I clustered the AP1 positive and HIF1 α bound peaks into GO Categories (biological processes). The significantly enriched categories were predominantly associated with developmental functions and apoptosis (Figure 26A). The main developmental categories that I found were angiogenesis and hematopoiesis. The mostly anticipated HIF1 α-associated cluster, glycolysis, was not significantly enriched. Additionally, the BioCarta graphical database revealed significant association with terminal differentiation of macrophages in the Raw.264 dataset (Figure 26B). About half of the genes required for this process were bound by HIF1 α in Raw.264 cells and at the same time contained AP1 binding sites. Together, these results demonstrate an important co-regulatory function of AP1 and Hif1 α. Co-regulation might be limited to genes with developmental and angiogenic functions and preferably occur at enhancer regions far away of the TSS.

Page 68 Results

Figure 25: AP1 transcription factors are adjacent to HREs. (A) Raw.264 HIF1 α ChIP-Seq data were ranked according to their score and clustered into 50 peak bins. Three highly overrepresented motifs within the ChIP-Seq dataset of Raw.264 cells were computed by Gibbs-Motif-Sampler. The average fold overrepresentation compared to an LSC sequence set was computed and plotted against the ranked bins. The green line represents the overrepresentation of a tandem HRE motif, which was one of the top motifs given by Gibbs-motif-Sampler. The black line represents a third order polynomial regression curve through each dataset. (B) In order to visualize the TRE distribution, each peak of each dataset was plotted sorted according to the distance to the TSS against the genomic location. The genomic region is represented by a black line, whereas a peak region is represented by a green line. A TRE site is represented as a blue spot and red if it is located within a peak region. Once a peak is harboring a TRE site, the green line is represented in yellow instead.

Page 69 Results

Figure 26: HIF1 ααα and AP1 co-regulated genes that were associated with development and differentiation. The TRE positive complete Group II-A ChIP-Seq data of Raw.264 cells (in total 917 genes), were analyzed by DAVID (A) and BioCarta (B) and clustered into GO categories o f biological processes. In (A) the top 20 overrepresented clusters are presented along with the respective count of Genes, pvalue, fold enrichment within the cluster and False Discovery Rate as computed by DAVID. In (B) the only significantly (pvalue < 0.0078) overrepresented pathway of the analysis, terminal differentiation of macrophages, is shown as an illustration. All bound target genes are marked by a red star.

Page 70 Results

9.6.3. The transcription factors SP1 and AP2 are candidates that might regulate hypoxia-induced genes in PMH independently of HIF1 α

By far the largest part of promoters of differentially expressed genes was not found to be bound by HIF1 α in primary cells. Therefore, promoters of differentially expressed genes that were not associated with HIF1 α binding were analyzed and screened for overrepresented motifs of known transcription factors. The analysis of the promoter region from -950 to +50bp with regard to the TSS was done by PScan (Zambelli et al., 2009) using Transfac as a library of known position weight matrices of transcription factors. In total, 1169 promoters of genes that were found to be more than 1.5 fold up regulated in PMH, were screened for overrepresented motifs. The most overrepresented motif was a -GC- rich box of the transcription factor SP1 (p-value < 1.16615e-27 for Transfac consensus V$SP1_Q6 and p-value < 3.49648e-22 for Transfac consensus V$SP1_01). The second strongly overrepresented consensus site was a binding site of the transcription factor family AP2 (p-value < 9.01831e-22 for Transfac consensus V$AP2_Q6). I subsequently assessed whether candidate transcription factors were induced upon hypoxia and whether they are direct Hif1 α targets. On the transcriptional level, SP1 showed a minor up-regulation after 3h of hypoxia of 1.31 fold and a moderate induction of 1.65 fold after 16h of hypoxia in PMH. However, the only significant binding event of HIF1 α was found within the Raw.264 dataset. For AP2, only AP2 delta showed a significant binding event in PMM and on the transcriptional level none of the tested cell types showed expression of AP2 isoforms. These data suggest a possible role of SP1 and AP2 in the regulation of genes that are differentially expressed upon long term hypoxia.

Page 71 Results

9.7. Downstream regulatory mechanism regulated by Hif1 ααα

9.7.1. The Hif1 α target JMJD3 mediates chromatin remodeling at the ADM promoter upon hypoxia

I next screened datasets for overrepresented GO categories that were involved in transcriptional regulation. A major class of genes, which is directly targeted and differentially expressed in all tested cells, is the JmjC domain containing protein class (Table 5). In Raw.264, more than half of all known JmjC domain-containing proteins were significantly bound at the PP region. Also within the 176 significant binding events of the PMM, six JmjC domain-containing proteins could be associated. However, at the expression level, only the PMH dataset showed a marked differential expression of many of the various JmjC domain-containing proteins.

Table 5: JMJC domain containing proteins are master regulators of the hypoxic response. All known JmjC domain-containing genes are represented in a table. For each dataset the rank within the ChIP datasets are provided and, if differentially expressed, the fold up regulation is provided.

Page 72 Results

Figure 27 : JMJD3 is enhanced on the transcriptional and on the protein level in PMH upon hypoxia. (A) RT-PCR was performed of total mRNA from PMH after different time points of exposure to hypoxic conditions (0.5% O 2). Primer pairs of three genes encoding JmjC do main containing proteins were chosen (JMJD1A, JMJD2B and JMJD3). (B) Western Blot analysis with 150 µg of nuclear extracts from PMH. Protein levels of HIF1 α were traced under normoxic conditions and after four time points of exposure to 0.5% oxygen. Lamin A was used as a control. Relative protein levels of JMJD3 as calculated by densitometric analysis are provided.

In order to assess the functional impact of HIF1 α binding to JmjC domain containing proteins, a targeted approach for enzyme activity was performed with one candidate of the bound histone demethylases, JMJD3. First, it was shown that JMJD3 is induced on mRNA level in PMH by more than two fold over the time course of hypoxia (Figure 27A). On the protein level JMJD3 accumulates approximately after 3h of hypoxic exposure and peaks after 16h with a 3.5 fold induction compared to Lamin A (Figure 27B). JMJD3 is demethylating H3K27me3, which is a Polycomb mediated repressor mark. I next sought for a classical target of Polycomb and JMJD3 that was induced and bound by Hif1 α upon hypoxia in PMH and then assessed its chromatin status. Adrenomedullin (ADM) was one candidate that fulfilled these criteria. I analyzed the whole promoter region for H3K27me3 using ChIP-QPCR. The significant binding event of HIF1 α occurs - 880bp in front of the ADM promoter (pvalue < 0.0004). In PMM no binding was observed at the ADM promoter region, however, in Raw.264 cells, a highly significant binding event was observed at the 3’ end of the gene (Figure 28A).

Page 73 Results

Under normoxic conditions, a peak for H3K27me3 was observed at around 500bp downstream of the TSS of ADM. After 3h of hypoxia this peak was decreasing

Figure 28 : JMJD3 is actively derepressing the ADM promoter upon exposure to hypoxia in PMH. (A) The ChIP-Seq results of the gene ADM within the Raw.264 cell dataset were plotted according to its genomic location. The black line as well as the heat plot represents the Hypoxia ChIP-Seq reads and the black line represents the Normoxia HIF1 α ChIP-Seq experiment. Conservation scores as well as genomic region are shown below. (B) ChIP-QPCR was performed with PMH and primer pairs designed to cover most of the PP region of ADM were used. The fold enrichment of the H3K27me3 mark after different t ime points of exposure to hypoxic conditions was plotted against the genomic location. (C) RT -PCR of ADM was performed of total mRNA from PMH after different time points of exposure to hypoxic conditions (0.5% O 2). background levels and after 16h of hypoxia the whole promoter was devoid of H3K27me3 marks (Figure 28B). Transcriptionally, ADM was found 16 times induced after 16h of exposure of the PMH cells to hypoxic conditions whereas after 3h only a minor induction (1.8 fold) occurred (Figure 28C).

Page 74 Results

It was previously shown that LPS induces JMJD3 through the transcription factor NF κB. Data in Raw.264 cells show a significant binding event of Hif1 α to a region approximately 5000bp upstream of the TSS (Figure 29A). In order to check whether HIF1 α induces, or contributes to induction of JMJD3, protein levels were assessed in Raw.264 cells after 2 and 8h of hypoxic exposure (Figure 29B). Hypoxia induced JMJD3 protein abundance 5.2 fold after 2h and 6.9 fold after 8h of exposure to hypoxia. However, the significance of this result has to be taken with care as this potentiating can only be seen after normalizing to Lamin A that markedly decreased upon hypoxia. Conclusively, these data suggest that JMJD3 is bound and regulated by Hif1 α and mediates demethylation of H3K27me3 at the ADM promoter upon hypoxia.

Page 75 Results

Figure 29: JMJD3 is enhanced on the protein level and bound by HIF1 ααα in Raw.264 cells. (A) The ChIP-Seq results of the gene JMJD3 within the Raw.264 cell dataset were plotted according to its genomic location. The black line as well as the heat plot represents the Hypoxia ChIP-Seq reads and the black line represents the Normoxia HIF1 α ChIP-Seq experiment. Conservation scores as well as genomic region are shown below. (B) Western Blot analysis with 150 µg of nuclear extracts from Raw.264 cells. Protein levels of HIF1a were traced under normoxic conditions and after two time points of exposure to 0.5% oxygen. Lamin A was used as a control. Relative protein levels of HIF1 α and JMJD3 as calculated by densitometric analysis are provided.

Page 76 Discussion

10. Discussion

It has been a major scientific goal to characterize HIF1 α-mediated transcriptional responses and to assign global functions to HIF1 α in the past. However, a comprehensive view on important transcriptional characteristics like cell specific binding and cooperative regulation with other transcription factors was still lacking. Here I present my integrative approach using genome wide binding and expression studies to analyze HIF1 α-mediated transcriptional responses to hypoxia.

10.1. Binding of HIF1 ααα is cell type specific

With primary mouse macrophages and primary hepatocytes, two cell types were used in this study that employ HIF1 α to adopt to low oxygen levels as previously shown (Cramer et al., 2003; Kim et al., 2006). To complement and validate the promoter-biased data of the ChIP-chip approach of PMM with a genome-wide approach, ChIP-Seq was performed using the murine leukaemic monocyte macrophage cell line, Raw.264. Comparison of the amount of binding events in both cell types revealed that HIF1 α binding is cell-type specific with only ~25% of overlap between PMM and PMH. The only genome-wide binding study comparing a transcription factor in different primary cells was described by (Odom et al., 2004). The study was performed with three transcription factors of the HNF family focusing only on PP regions. It revealed an overlap of approximately 50% of binding events between primary, pancreatic beta islets and primary hepatocytes. Therefore, despite using a FDRs of 20% and 14% for PMH and PMM, respectively, the overlap in these cell types is about half of the one in the HNF study indicating that HIF binding is strongly cell type-dependent.

Page 77 Discussion

One reason for this may be that the epigenetic status and maintenance by histone modifiers is markedly affected upon exposure to hypoxia and only few differentially expressed factors in PMM compared to PMH can change the transcriptional pattern of the cell (chapter 6.2.1). Indeed, several histone modifiers of the JmjC family were bound and differentially expressed between the two cell types (Table 5) and the cell type- specific methylation status of the DNA may further contribute to the cell type-specific accessibility of HIF1 α target sites. However, the dynamics of binding events was not addressed during a time course and it could well be that HIF1 α promoter occupation occurs at different time points in different cell types. Finally also technical limitations should be considered since the ChIP-chip approach involves several critical steps until a final binding pattern can be proposed. First the purity of the primary cell populations can vary among replicates. Thioglycollate- elicited macrophages might be contaminated with an unknown proportion of other cell types of the myeloid lineage (e.g. neutrophils) and hepatocytes elicited by liver perfusion may contain Kupffer cells. Additionally, although stringently monitored, I observed unavoidable differences in cross-linking- and shearing efficiency in the two cell types.

10.2. One out of five genes that are bound by HIF1 ααα are differentially expressed in PMH and PMM

Integration of the expression data after 16h of hypoxia into the binding dataset at 3h of hypoxia revealed that 19% and 25% of the genes that are bound in PMH and PMM respectively, are also significantly up-regulated. This finding is well in line with previous studies integrating binding into expression data. In fact, only one ChIP-chip study with FoxP3 (Zheng et al., 2007) was showing a higher overlap between both datasets. The overlap of down-regulated genes was about five to ten times lower, which is in line with previous studies in which HIF1 α has been demonstrated to constitute a transcriptional activator rather than a repressor.

Page 78 Discussion

It is important to mention that a stringent cut off of raw intensities of 500 was chosen to exclude genes that are only marginally expressed. This set-up avoids false positive differentially expressed genes due to the lack of sensitivity of the arrays. On the other hand, low sensitivity and limitations in the design of ESTs in these arrays may also contribute to false negative results. In any case, genes which passed these criteria were considered to be actively transcribed. For both cell types, these were about 50% of genes spotted on the whole genome expression arrays. Interestingly, if only binding events of genes that are actively transcribed, are considered for comparison to the expression data, the overlap of differentially expressed and bound genes can be increased to more than 60% between PMM and PMH. Technically, this could be due to the limitations in sensitivity and design of the expression array. Another ex0planation could be that probably many genes bound by HIF1 α are missing important prerequisites for active transcription such as DNA and histone modifications as well as co-factor recruitment. The limited overlap between PMM and PMH also suggests that these differences may be cell type specific.

10.3. One out of twenty-five hypoxia responsive genes are bound by HIF1 ααα in PMH

In total, about 8 and 25 times more genes were significantly up regulated upon exposure to hypoxia than bound by HIF1 α in PMM and PMH, respectively suggesting that most of the transcriptional changes induced by hypoxia occurred either secondary to- or independent of HIF1 α. Since physiological changes upon adaptation to hypoxic conditions are profound (chapter 6.1.1), a significant change of the expression pattern in the hypoxic cell is expected. These changes in response to cellular stress can induce other transcription factors such as for example AP1. Furthermore, the epigenetic pattern is dramatically changing during hypoxia (chapter 6.2.3). My work proposes several candidates for secondary regulation of differentially expressed genes (see below) These candidates however need to be confirmed in the future.

Page 79 Discussion

Another interpretation of the high amount of differentially expressed genes that are not directly targeted by HIF1 α can be the highly dynamic process of transcriptional regulation. Although, we and others (Ohnishi et al., 2007; Xia et al., 2009) showed that HIF1 α binding is best reflected by differential expression after 12-24 hours of exposure to 0.5% oxygen, several genes might have been differentially expressed at early time points after HIF1 α binding, whereas others are differentially expressed upon long term hypoxia only. Therefore, to address the regulatory impact of HIF1 α binding, a high resolution assessment of HIF1 α binding during a time course would have to be integrated into a high resolution time-dependent analysis of transcript levels upon hypoxia.

10.4. ChIP-Seq reveals markedly more HIF1 ααα binding events during hypoxia in Raw.264 cells

Using a genome-wide ChIP-Seq approach in Raw.264 cells, a more than 10 fold increase in binding events within the PP could be detected compared to ChIP-chip in primary cells. Many previously published genome-wide binding studies revealed significant binding events of transcription factors in the amount of thousands to ten thousands (Table 1). However, two recent genome wide binding studies with HIF1 α using ChIP-chip proposed significant binding of about 150 – 600 genes. The reason for the enormous difference in the amounts of direct HIF1 α targets between our or others’ ChIP-chip-based association studies and our ChIP-Seq-derived data can be explained from a technical or biological point of view. The technical difference can be due to the different methods used for binding site detection after ChIP. It has been demonstrated that ChIP-Seq provides an enhanced sensitivity as well as an increased resolution compared to ChIP-chip used in the other studies (Kharchenko et al., 2008), the differences can be well explained. On the other hand, it can be speculated that HIF1 α levels are enhanced in Raw.264 cells compared to primary cells and other cell types used in the respective studies. Indeed, it has been shown, that increased abundance of transcription factors is correlating with increased binding. However, to exclude one or the other hypothesis, further experiments would have to be demonstrated.

Page 80 Discussion

10.5. HIF1 ααα directly binds to genes associated to glycolysis, angiogenesis and regulation of transcription, depending on the cell type.

Functional clustering of the binding events in both cell types revealed striking similarities as well as differences. Among all tested datasets, glycolysis was the top overrepresented cluster. However, unlike PMM, PMH showed an overrepresentation of genes involved in angiogenesis and blood vessel development. The two cell type specific clusters in PMMs were anion transport and general, DNA-dependent transcription. Similar analysis using the top 500 genes bound in Raw.264 cells confirmed the enrichment for genes involved in regulation of transcription. Therefore it can be speculated that hepatocytes more likely react to low oxygen levels by transcription of angiogenic factors compared to macrophages. In macrophages, HIF1 α seems to specifically induce anion transporters. This would be an important contribution of HIF1 α in this cell type, in order to maintain physiological pH levels under hypoxia. It has been shown that macrophages are relying on anaerobic glycolysis to maintain energy levels under hypoxia and even upon normoxic conditions (Cramer et al., 2003), the end product of anaerobic glycolysis, lactate, which is in fact a natural occurring anion has to be removed constantly. However, it has to be noted, that the interpretation of biological processes, overrepresented in a set of genes has to be taken with care, since one major factor of a cluster can be biologically more meaningful than a whole battery of genes overrepresented and assigned to the same cluster (Rhee et al., 2008).

10.6. HIF1 ααα preferentially binds close to the TSS

Assessing localization of binding events, I showed a strong bias towards the TSS. Approximately more than 60% of the binding events within the PP occurred in close proximity to the TSS (+-1kbp) in all tested cell types. The fraction of peaks located within the +-1kbp window is more than 35% of all group IIB peaks after using an extended window of +-100kbp around the TSS in the whole genome Raw.264 cell dataset. This finding confirms the observations of other genome wide binding studies of other transcription factors (Xia et al., 2009).

Page 81 Discussion

This effect was even more pronounced by analysis of the group of genes that were differentially expressed in each primary cell type. The percentages of differentially expressed genes with a HIF1 α binding event within the first 1kbp surrounding the TSS, were approximately 2 times higher than the ones located 1kbp more distant to the TSS. One interpretation of these results may be that transcription factor binding to the TSS is directly linked to transcriptional initiation in response to hypoxia, while binding to enhancer regions might have some modulatory effects that are not necessarily dependent on hypoxia. It has to be explored in the future what the physiologic triggers of enhancer binding might be and what the consequences of these modifications might be. A similar effect was seen if peaks where filtered for HRE containing sequences in all tested cell types. The fraction of genes that were bound by HIF1 α lacking an HRE showed decreased binding scores and a localization more distant to the TSS. The lower binding scores seen in the HRE negative fraction of peaks might either underscore the decreased statistical reliability of these peaks or the decreased affinity of the transcriptional complex involving HIF1 α. Apart from the bias to the TSS no preferred binding site could be located. However, an overrepresentation of a characteristic binding pattern, which could be observed simultaneously at the 5´- and at the 3´ end was found by analysis of the Raw.264 dataset. This suggests a common mechanism of transcriptional regulation by HIF1 α, however the functional impact of this binding pattern has to be determined by expression analysis.

10.7. HIF1 ααα preferentially binds to an HRE consisting of nine base pairs or an tandem core HRE

Using a well-established motif search algorithm for mammalian sequences, all peak sequences were analyzed in order to characterize the motif targeted by HIF1 α. The resulting data strongly indicate that HIF1 α favorably targets a 9bp long motif (5´- CGTACGTGC-3´), 4bp longer than the established core HRE. Furthermore a fraction of approximately 15% of the significant peak sequences of all datasets showed a strong overrepresentation of tandem HRE (5´-CGTGNNNNNCGTG- 3´), consisting of two core HREs (5´-CGTG-3´) and a nonspecific, 5bp long intervening sequence.

Page 82 Discussion

Together these data suggest that HIF1 α favors CpG rich consensus sites. This enables the cell to persistently modulate HIF1 α binding by methylation of the CpG (Wenger et al., 2005). Furthermore, HIF1 α seems to bind cooperatively to promoters by preferable binding to tandem HREs. However, it has to be shown whether two molecules of HIF1 α can bind simultaneously to each core HRE within the tandem HRE.

10.8. The TRE consensus motif is overrepresented at enhancer regions targeted by HIF1 ααα

Further sequence analysis using a different algorithm that filters for redundant motifs, revealed a strong and consistent overrepresentation of AP1 consensus sites. I could show that these motifs are preferentially found at peak sequences far away from the TSS in Raw.264 cells. Gene ontological clustering for biological processes of the genes associated to these AP1 positive peaks revealed a strong overrepresentation in clusters associated to small GTPase-mediated signal transduction, blood vessel development and apoptosis, which are among the anticipated ontological clusters regulated by AP1 transcription factors (Eychene et al., 2008; Jochum et al., 2001). The enrichment of AP1 consensus motifs may be either due to indirect enrichment through long distance interactions with the HIF1 α associated transcriptional complex or by direct interaction of HIF1 α and AP1 at enhancer sites. However it is difficult to estimate the specific factor of the AP1 family that might be a primary candidate for this interaction since AP1 transcription factors are composed heterodimers belonging to the c-Fos, c-Jun, ATF and JDP families. Furthermore AP1 factors are induced by a wide range of stimuli e.g. cytokines, growth factors, stress, and bacterial and viral infections (Eferl and Wagner, 2003; Shaulian and Karin, 2001). Therefore, conclusively it can be suggested that HIF1 α might be not sufficient to regulate the expression of the genes associated to enriched sequences by HIF1 α ChIP harboring an AP1 target site. Since both transcription factors are activator transcription factors, the presence of AP1 factors might be necessary to modulate developmental and apoptotic functions upon hypoxia.

Page 83 Discussion

Furthermore clustering of the TRE positive HIF1 α bound genes in gene ontological categories by the BioCarta graphical database revealed a significant association to the process of terminal macrophage differentiation. Especially the two ETS transcription factors, ETS1 and ETS2 which are known to interact with HIF1 α (Salnikow et al., 2008), were among these genes and are involved in a variety of differentiation processes such as hematopoiesis (Sharrocks, 2001). Since developmental processes often take place at sites of low oxygen levels (chapter 6.1.5), it might well be that HIF1 α, together with AP1 family members induce ETS transcription factors in order to regulate the progression of cellular differentiation.

10.9. SP1 is a potential HIF1 ααα target and might regulate genes in response to hypoxia independent of HIF1 ααα

Using the same algorithm that filters for redundant motifs, but analyzing only the promoters of genes, which are not bound, but differentially expressed, revealed overrepresentation of a consensus motif targeted by families of two well described transcription factors, SP1 and AP2. The ubiquitous transcription factor SP1 is known to constitutively activate housekeeping genes lacking a classical TATA-Box but was recently also associated to a variety of other processes such as differentiation (Wierstra, 2008). Genes that were found differentially expressed but not bound by HIF1 α upon hypoxia may be regulated by SP1 since the latter transcription factor is transcriptionally induced in response to hypoxia. However, since no binding of HIF1 α to SP1 could be observed, this induction is either secondary or even HIF-independent. Therefore SP1 might be induced by hypoxia and subsequently enhance expression of housekeeping genes lacking a TATA-Box and leading to a differential expression of an unknown subset of genes measured by expression analysis in PMH upon 16h of hypoxia.

Page 84 Discussion

The second motif overrepresented on the promoters of genes differentially expressed but not bound by HIF1 α under hypoxia was the AP2 consensus motif. The transcription factors of the AP2 family are mainly associated to regulation of developmental processes (Eckert et al., 2005). Although not expressed in PMM and PMH, AP2 might be of important function in other cell types to further enhance expression of developmental genes upon exposure to hypoxia independently of HIF1 α.

10.10. Transcriptional regulation of chromatin modifiers by HIF1 ααα

An intriguing link of HIF1 α to the regulation of chromatin modifiers of the JmjC family was previously reported, however the functional impact of this link has been questioned by the authors due to the common assumption that histone demethylation is oxygen- dependent (Xia et al., 2009). Although a recent study showed decreased methylation under hypoxia, the mechanism of this process is currently unknown (chapter 6.2.3). I could confirm this link by showing that HIF1 α binds to several JmjC family members in all tested cell types. Moreover, I tested one member of the JmjC family, JMJD3, which is significantly bound in PMM and Raw.264 cells and almost significantly bound in PMH and differentially expressed in all tested cell types, for its functional impact during hypoxia. The results showed, that the promoter of ADM, a gene that is targeted by HIF1 α only at the 3´end, clearly loses its repressive H3K27me3 mark during the course of hypoxia in PMH and this loss strongly correlates with the up regulation of the ADM transcript. Together, these results suggests an important link between the dynamic, epigenetic regulation of genes by JMJD3 and possibly other demethylases, and transcriptional changes observed under hypoxia. A similar link has been previously shown specifically for JMJD3 in response to inflammatory stimuli (De Santa et al., 2007).

Page 85 Discussion

10.11. Comparison to previous genome wide HIF1 ααα binding studies

Two very recent studies discovered novel functions of HIF1 α by ChIP-chip using DNA tiling arrays in combination with expression analysis. Kung et al utilized HepG2, a human hepatoma cell line, for a genome wide binding and expression study and integrated expression data of U87 and MBA-231 cells, a human glioblastoma and a human breast cancer cell line respectively, into the HepG2 data set (Xia et al., 2009). The data revealed that 50% of a total of 283 detected significant binding events were linked to promoters of known genes. Interestingly and in line with my study, four JmjC domain containing proteins were found to be direct targets and 17 JmjC family members had significantly increased mRNA abundance under hypoxic conditions. One JmjC family member, Jarid1B, was functionally assessed. It was concluded, that induction of JmjC family members under hypoxia was to compensate for their lower enzymatic activity at low oxygen levels and thus to maintain the methylation levels found at normoxic conditions which is contradictory to the data I observed by analysis of JMJD3 (chapter 9.7.1). Furthermore, analysis with two de novo motif search algorithms, including Weeder, also used in this thesis, revealed the classical HRE (5’-RCGTG-3’) as the most overrepresented motif in the dataset without any preference to extended motifs. However, it has to be noted that according to internal recalculations with the methods used in this thesis, significantly less genes could be associated to the peaks (less than 150 in the PP) detected in this study. This was presumably due to filtering for duplicate peak-gene associations and a more stringent gene definition used in my dataset. The overlap between this dataset to my data was only about 15% to Raw.264 cells, PMM and PMH.

Page 86 Discussion

The second binding study of HIF1 α and HIF2a using MCF7 cells, a human breast cancer-derived cell line, was a comparative study using promoter arrays (sequences from -7.5kbp to +2.5kbp with regard to the TSS). In total 546 and 143 binding events could be mapped to 394 and 134 genes in case of HIF1 α and HIF2a, respectively. The overlap of the HIF2a data to my data were marginally; the overlap of the MCF7 HIF1 α ChIP-chip data to my dataset were less than 20% compared to Raw.264 cells, PMM and PMH. As outlined above, only the classical HRE (5’-RCGTG-3’) was found to be the most overrepresented motif in the dataset without any preference to extended motifs. Subsequent comparison of the binding data to the expression data revealed that only 20.8% of the HIF1 α bound genes were at the same time differentially expressed by exposure to either Hypoxia or DMOG. Moreover, siRNA-directed knockdown of HIF2a revealed, that only one out of ninety differentially expressed genes that were bound by both isoforms lost its differential expression suggesting HIF1 α to be the crucial isoform to regulate hypoxia induced transcriptional responses. Together, these studies revealed and confirmed the finding in primary cells, that HIF1 α binds approximately to 150-400 genes, depending on the cell type. However, the overlaps between all studies are not exceeding 20%, highlighting the cell type specific function of HIF1 α

Page 87 Outlook

11. Outlook

Using proximal promoter arrays, the HIF1 α binding data in PMH and PMM suggested that HIF1 α binds in a cell type-specific manner. This conclusion was based on the finding of relatively low overlaps (~25%). It would be interesting to see whether ChIP- Seq derived data - by combining it with high resolution expression analysis - would be equally different assuming the higher resolution and specificity of this method. Target motifs could be refined and genome wide observations, far from the promoter region, could be provided. Moreover, Agilent genome wide expression arrays are covering only 1 to 3 exons per gene. With integration of RNA-Seq, a method that uses sequencing to analyze genome wide transcript levels and therefore accounts for all spliced isoforms expressed, a genome wide comparison between different cell types could be performed. The genome wide location study of HIF1 α in Raw.264 cells revealed several intriguing results including binding patterns, motif refinements and proposal of new cooperating candidate transcription factors. However, due to the lack of expression data, no functional information was provided, following HIF1α binding. Therefore, a RNA-Seq experiment would extend the view on the current findings. Additionally the huge amount of bound target genes in Raw.264 cells was measured by one ChIP-Seq experiment. To confirm the data at different score groups and at different loci, ChIP-QPCR will be performed in the near future. In addition, a solid FDR could be estimated for this dataset. Analysing the sequences, an extended 9bp core HRE could be proposed. To determine, whether HIF1 α shows an increased affinity to this site, luciferase assays should be performed. Furthermore, it was speculated that HIF1 α might interact with AP1 transcription factors in order to regulate developmental genes by interaction at enhancer sites. However, until now it is not clear whether this interaction is cooperative or due to long-range interaction. Therefore, the TRE+ HIF1 α bound sequences should be further analyzed and ChIP- QPCR experiments with HIF1 α and ideally, AP1 members should reveal more of the characteristics of this interaction.

Page 88 Outlook

Furthermore, it would be interesting to see how expression levels of selected HIF1 α TRE+ genes behave on the transcriptional level if either HIF1 α and / or AP1 member are inactivated. A limited siRNA screen with AP1 members, tracing a few HIF1 α bound TRE+ candidate transcripts, would give a good insight at the significance of our hypothesis. Similar approaches could be followed to study the importance of SP1 and AP2 members in order to differentially express genes that are not directly bound by HIF1 α. Therefore, a knockdown of the AP2 members or the SP1 members and subsequent measurments of transcript levels of genes being associated to the respective consensus site during hypoxia should reveal the significance of these factors under hypoxic conditions. Finally yet importantly the unexpected functional activity of JMJD3 under hypoxia provided an intriguing link between hypoxia and dynamic epigenetic changes mediated by JmjC family members. However, several follow up experiments have to be performed to validate these results. First of all, knock-down of either HIF1 α or JMJD3 in response to hypoxia should abolish the demethylation of H3K27me3. Furthermore, more promoters with known Polycomb interaction and known induction under hypoxic conditions have to be tested for H3k27me3 levels. A ChIP-Seq experiment of JMJD3 and H3K27me3 under normoxia compared to long-term hypoxia and, ideally, RNA-Seq studies should complete the picture of this unexpected finding.

Page 89 References

12. References

Arnett-Mansfield, R.L., Graham, J.D., Hanson, A.R., Mote, P.A., Gompel, A., Scurr, L.L., Gava, N., de Fazio, A., and Clarke, C.L. (2007). Focal subnuclear distribution of progesterone receptor is ligand dependent and associated with transcriptional activity. Mol Endocrinol 21 , 14-29. Belaiba, R.S., Bonello, S., Zahringer, C., Schmidt, S., Hess, J., Kietzmann, T., and Gorlach, A. (2007). Hypoxia up-regulates hypoxia-inducible factor-1alpha transcription by involving phosphatidylinositol 3-kinase and nuclear factor kappaB in pulmonary artery smooth muscle cells. Mol Biol Cell 18 , 4691-4697. Benhamouche, S., Decaens, T., Godard, C., Chambrey, R., Rickman, D.S., Moinard, C., Vasseur-Cognet, M., Kuo, C.J., Kahn, A., Perret, C. , et al. (2006). Apc tumor suppressor gene is the "zonation-keeper" of mouse liver. Dev Cell 10 , 759-770. Berger, S.L. (2007). The complex language of chromatin regulation during transcription. Nature 447 , 407-412. Berta, M.A., Mazure, N., Hattab, M., Pouyssegur, J., and Brahimi-Horn, M.C. (2007). SUMOylation of hypoxia-inducible factor-1alpha reduces its transcriptional activity. Biochem Biophys Res Commun 360 , 646-652. Bjornheden, T., Levin, M., Evaldsson, M., and Wiklund, O. (1999). Evidence of hypoxic areas within the arterial wall in vivo. Arterioscler Thromb Vasc Biol 19 , 870-876. Braeuning, A., Ittrich, C., Kohle, C., Hailfinger, S., Bonin, M., Buchmann, A., and Schwarz, M. (2006). Differential gene expression in periportal and perivenous mouse hepatocytes. FEBS J 273 , 5051-5061. Brahimi-Horn, M.C., and Pouyssegur, J. (2009). HIF at a glance. J Cell Sci 122 , 1055- 1057. Caldwell, C.C., Kojima, H., Lukashev, D., Armstrong, J., Farber, M., Apasov, S.G., and Sitkovsky, M.V. (2001). Differential effects of physiologically relevant hypoxic conditions on T lymphocyte development and effector functions. J Immunol 167 , 6140-6149. Carmeliet, P., Dor, Y., Herbert, J.M., Fukumura, D., Brusselmans, K., Dewerchin, M., Neeman, M., Bono, F., Abramovitch, R., Maxwell, P. , et al. (1998). Role of HIF- 1alpha in hypoxia-mediated apoptosis, cell proliferation and tumour angiogenesis. Nature 394 , 485-490. Carninci, P., Kasukawa, T., Katayama, S., Gough, J., Frith, M.C., Maeda, N., Oyama, R., Ravasi, T., Lenhard, B., Wells, C. , et al. (2005). The transcriptional landscape of the mammalian genome. Science 309 , 1559-1563. Carroll, J.S., Meyer, C.A., Song, J., Li, W., Geistlinger, T.R., Eeckhoute, J., Brodsky, A.S., Keeton, E.K., Fertuck, K.C., Hall, G.F. , et al. (2006). Genome-wide analysis of estrogen receptor binding sites. Nat Genet 38 , 1289-1297. Ceradini, D.J., Kulkarni, A.R., Callaghan, M.J., Tepper, O.M., Bastidas, N., Kleinman, M.E., Capla, J.M., Galiano, R.D., Levine, J.P., and Gurtner, G.C. (2004). Progenitor cell trafficking is regulated by hypoxic gradients through HIF-1 induction of SDF-1. Nat Med 10 , 858-864. Chen, H., Yan, Y., Davidson, T.L., Shinkai, Y., and Costa, M. (2006). Hypoxic stress induces dimethylated histone H3 lysine 9 through histone methyltransferase G9a in mammalian cells. Cancer Res 66 , 9009-9016.

Page 90 References

Chen, X., Xu, H., Yuan, P., Fang, F., Huss, M., Vega, V.B., Wong, E., Orlov, Y.L., Zhang, W., Jiang, J. , et al. (2008). Integration of external signaling pathways with the core transcriptional network in embryonic stem cells. Cell 133 , 1106- 1117. Chiche, J., Ilc, K., Laferriere, J., Trottier, E., Dayan, F., Mazure, N.M., Brahimi-Horn, M.C., and Pouyssegur, J. (2009). Hypoxia-inducible carbonic anhydrase IX and XII promote tumor cell growth by counteracting acidosis through the regulation of the intracellular pH. Cancer Res 69 , 358-368. Cloos, P.A., Christensen, J., Agger, K., and Helin, K. (2008). Erasing the methyl mark: histone demethylases at the center of cellular differentiation and disease. Genes Dev 22 , 1115-1140. Cole, M.F., Johnstone, S.E., Newman, J.J., Kagey, M.H., and Young, R.A. (2008). Tcf3 is an integral component of the core regulatory circuitry of embryonic stem cells. Genes Dev 22 , 746-755. Cramer, T., Yamanishi, Y., Clausen, B.E., Forster, I., Pawlinski, R., Mackman, N., Haase, V.H., Jaenisch, R., Corr, M., Nizet, V. , et al. (2003). HIF-1alpha is essential for myeloid cell-mediated inflammation. Cell 112 , 645-657. Crawford, T.N., Alfaro, D.V., 3rd, Kerrison, J.B., and Jablon, E.P. (2009). Diabetic retinopathy and angiogenesis. Curr Diabetes Rev 5, 8-13. De Santa, F., Totaro, M.G., Prosperini, E., Notarbartolo, S., Testa, G., and Natoli, G. (2007). The histone H3 lysine-27 demethylase Jmjd3 links inflammation to inhibition of polycomb-mediated gene silencing. Cell 130 , 1083-1094. Dekker, J. (2008). Gene regulation in the third dimension. Science 319 , 1793-1794. Denko, N.C. (2008). Hypoxia, HIF1 and glucose metabolism in the solid tumour. Nat Rev Cancer. Depping, R., Steinhoff, A., Schindler, S.G., Friedrich, B., Fagerlund, R., Metzen, E., Hartmann, E., and Kohler, M. (2008). Nuclear translocation of hypoxia-inducible factors (HIFs): involvement of the classical importin alpha/beta pathway. Biochim Biophys Acta 1783 , 394-404. Dufour, C.R., Wilson, B.J., Huss, J.M., Kelly, D.P., Alaynick, W.A., Downes, M., Evans, R.M., Blanchette, M., and Giguere, V. (2007). Genome-wide orchestration of cardiac functions by the orphan nuclear receptors ERRalpha and gamma. Cell Metab 5, 345-356. Eckert, D., Buhl, S., Weber, S., Jager, R., and Schorle, H. (2005). The AP-2 family of transcription factors. Genome Biol 6, 246. Eferl, R., and Wagner, E.F. (2003). AP-1: a double-edged sword in tumorigenesis. Nat Rev Cancer 3, 859-868. Elvidge, G.P., Glenny, L., Appelhoff, R.J., Ratcliffe, P.J., Ragoussis, J., and Gleadle, J.M. (2006). Concordant regulation of gene expression by hypoxia and 2- oxoglutarate-dependent dioxygenase inhibition: the role of HIF-1alpha, HIF- 2alpha, and other pathways. J Biol Chem 281 , 15215-15226. Eychene, A., Rocques, N., and Pouponnot, C. (2008). A new MAFia in cancer. Nat Rev Cancer 8, 683-693. Frede, S., Stockmann, C., Freitag, P., and Fandrey, J. (2006). Bacterial lipopolysaccharide induces HIF-1 activation in human monocytes via p44/42 MAPK and NF-kappaB. Biochem J 396 , 517-527. Gordan, J.D., Bertout, J.A., Hu, C.J., Diehl, J.A., and Simon, M.C. (2007). HIF-2alpha promotes hypoxic cell proliferation by enhancing c-myc transcriptional activity. Cancer Cell 11 , 335-347.

Page 91 References

Guelen, L., Pagie, L., Brasset, E., Meuleman, W., Faza, M.B., Talhout, W., Eussen, B.H., de Klein, A., Wessels, L., de Laat, W. , et al. (2008). Domain organization of human chromosomes revealed by mapping of nuclear lamina interactions. Nature 453 , 948-951. Guzy, R.D., and Schumacker, P.T. (2006). Oxygen sensing by mitochondria at complex III: the paradox of increased reactive oxygen species during hypoxia. Exp Physiol 91 , 807-819. Hamaguchi, T., Iizuka, N., Tsunedomi, R., Hamamoto, Y., Miyamoto, T., Iida, M., Tokuhisa, Y., Sakamoto, K., Takashima, M., Tamesa, T. , et al. (2008). Glycolysis module activated by hypoxia-inducible factor 1alpha is related to the aggressive phenotype of hepatocellular carcinoma. Int J Oncol 33 , 725-731. Harada, H., Itasaka, S., Kizaka-Kondoh, S., Shibuya, K., Morinibu, A., Shinomiya, K., and Hiraoka, M. (2009). The Akt/mTOR pathway assures the synthesis of HIF- 1alpha protein in a glucose- and reoxygenation-dependent manner in irradiated tumors. J Biol Chem 284 , 5332-5342. Henikoff, S., Strahl, B.D., and Warburton, P.E. (2008). Epigenomics: a roadmap to chromatin. Science 322 , 853. Herst, P.M., and Berridge, M.V. (2007). Cell surface oxygen consumption: a major contributor to cellular oxygen consumption in glycolytic cancer cell lines. Biochim Biophys Acta 1767 , 170-177. Hu, C.J., Sataur, A., Wang, L., Chen, H., and Simon, M.C. (2007). The N-terminal transactivation domain confers target gene specificity of hypoxia-inducible factors HIF-1alpha and HIF-2alpha. Mol Biol Cell 18 , 4528-4542. Iyer, N.V., Kotch, L.E., Agani, F., Leung, S.W., Laughner, E., Wenger, R.H., Gassmann, M., Gearhart, J.D., Lawler, A.M., Yu, A.Y. , et al. (1998). Cellular and developmental control of O2 homeostasis by hypoxia-inducible factor 1 alpha. Genes Dev 12 , 149-162. Jantsch, J., Chakravortty, D., Turza, N., Prechtel, A.T., Buchholz, B., Gerlach, R.G., Volke, M., Glasner, J., Warnecke, C., Wiesener, M.S. , et al. (2008). Hypoxia and hypoxia-inducible factor-1 alpha modulate lipopolysaccharide-induced dendritic cell activation and function. J Immunol 180 , 4697-4705. Jeong, J.W., Bae, M.K., Ahn, M.Y., Kim, S.H., Sohn, T.K., Bae, M.H., Yoo, M.A., Song, E.J., Lee, K.J., and Kim, K.W. (2002). Regulation and destabilization of HIF- 1alpha by ARD1-mediated acetylation. Cell 111 , 709-720. Ji, H., Jiang, H., Ma, W., Johnson, D.S., Myers, R.M., and Wong, W.H. (2008). An integrated software system for analyzing ChIP-chip and ChIP-seq data. Nat Biotechnol 26 , 1293-1300. Jochum, W., Passegue, E., and Wagner, E.F. (2001). AP-1 in mouse development and tumorigenesis. Oncogene 20 , 2401-2412. Johnson, A.B., Denko, N., and Barton, M.C. (2008a). Hypoxia induces a novel signature of chromatin modifications and global repression of transcription. Mutat Res 640 , 174-179. Johnson, D.S., Li, W., Gordon, D.B., Bhattacharjee, A., Curry, B., Ghosh, J., Brizuela, L., Carroll, J.S., Brown, M., Flicek, P. , et al. (2008b). Systematic evaluation of variability in ChIP-chip experiments using predefined DNA targets. Genome Res 18 , 393-403. Jungermann, K., and Kietzmann, T. (1996). Zonation of parenchymal and nonparenchymal metabolism in liver. Annu Rev Nutr 16 , 179-203. Kaelin, W.G., Jr., and Ratcliffe, P.J. (2008). Oxygen sensing by metazoans: the central role of the HIF hydroxylase pathway. Mol Cell 30 , 393-402.

Page 92 References

Kharchenko, P.V., Tolstorukov, M.Y., and Park, P.J. (2008). Design and analysis of ChIP-seq experiments for DNA-binding proteins. Nat Biotechnol 26 , 1351-1359. Kietzmann, T., Roth, U., and Jungermann, K. (1999). Induction of the plasminogen activator inhibitor-1 gene expression by mild hypoxia via a hypoxia response element binding the hypoxia-inducible factor-1 in rat hepatocytes. Blood 94 , 4177-4185. Kim, T.H., Abdullaev, Z.K., Smith, A.D., Ching, K.A., Loukinov, D.I., Green, R.D., Zhang, M.Q., Lobanenkov, V.V., and Ren, B. (2007). Analysis of the vertebrate insulator protein CTCF-binding sites in the human genome. Cell 128 , 1231- 1245. Kim, T.H., Barrera, L.O., Zheng, M., Qu, C., Singer, M.A., Richmond, T.A., Wu, Y., Green, R.D., and Ren, B. (2005). A high-resolution map of active promoters in the human genome. Nature 436 , 876-880. Kim, W.Y., Safran, M., Buckley, M.R., Ebert, B.L., Glickman, J., Bosenberg, M., Regan, M., and Kaelin, W.G., Jr. (2006). Failure to prolyl hydroxylate hypoxia- inducible factor alpha phenocopies VHL inactivation in vivo. EMBO J 25 , 4650- 4662. Kouzarides, T. (2007). Chromatin modifications and their function. Cell 128 , 693-705. Lando, D., Peet, D.J., Gorman, J.J., Whelan, D.A., Whitelaw, M.L., and Bruick, R.K. (2002). FIH-1 is an asparaginyl hydroxylase enzyme that regulates the transcriptional activity of hypoxia-inducible factor. Genes Dev 16 , 1466-1471. Levine, M., and Tjian, R. (2003). Transcription regulation and animal diversity. Nature 424 , 147-151. Li, B., Carey, M., and Workman, J.L. (2007a). The role of chromatin during transcription. Cell 128 , 707-719. Li, F., Sonveaux, P., Rabbani, Z.N., Liu, S., Yan, B., Huang, Q., Vujaskovic, Z., Dewhirst, M.W., and Li, C.Y. (2007b). Regulation of HIF-1alpha stability through S-nitrosylation. Mol Cell 26 , 63-74. Lin, Q., Kim, Y., Alarcon, R.M., and Yun, Z. (2008). Oxygen and Cell Fate Decisions. Gene Regul Syst Bio 2, 43-51. Linhart, C., Halperin, Y., and Shamir, R. (2008). Transcription factor and microRNA motif discovery: the Amadeus platform and a compendium of metazoan target sets. Genome Res 18 , 1180-1189. Lisy, K., and Peet, D.J. (2008). Turn me on: regulating HIF transcriptional activity. Cell Death Differ 15 , 642-649. Liu, L., Cash, T.P., Jones, R.G., Keith, B., Thompson, C.B., and Simon, M.C. (2006). Hypoxia-induced energy stress regulates mRNA translation and cell growth. Mol Cell 21 , 521-531. Makino, Y., Cao, R., Svensson, K., Bertilsson, G., Asman, M., Tanaka, H., Cao, Y., Berkenstam, A., and Poellinger, L. (2001). Inhibitory PAS domain protein is a negative regulator of hypoxia-inducible gene expression. Nature 414 , 550-554. Mantena, S.K., Vaughn, D.P., Andringa, K.K., Eccleston, H.B., King, A.L., Abrams, G.A., Doeller, J.E., Kraus, D.W., Darley-Usmar, V.M., and Bailey, S.M. (2009). High fat diet induces dysregulation of hepatic oxygen gradients and mitochondrial function in vivo. Biochem J 417 , 183-193. Mardis, E.R. (2008). Next-generation DNA sequencing methods. Annu Rev Genomics Hum Genet 9, 387-402. Margueron, R., Trojer, P., and Reinberg, D. (2005). The key to development: interpreting the histone code? Curr Opin Genet Dev 15 , 163-176.

Page 93 References

Martin, C., and Zhang, Y. (2005). The diverse functions of histone lysine methylation. Nat Rev Mol Cell Biol 6, 838-849. Nielsen, R., Pedersen, T.A., Hagenbeek, D., Moulos, P., Siersbaek, R., Megens, E., Denissov, S., Borgesen, M., Francoijs, K.J., Mandrup, S. , et al. (2008). Genome-wide profiling of PPARgamma:RXR and RNA polymerase II occupancy reveals temporal activation of distinct metabolic pathways and changes in RXR dimer composition during adipogenesis. Genes Dev 22 , 2953- 2967. Odom, D.T., Zizlsperger, N., Gordon, D.B., Bell, G.W., Rinaldi, N.J., Murray, H.L., Volkert, T.L., Schreiber, J., Rolfe, P.A., Gifford, D.K. , et al. (2004). Control of pancreas and liver gene expression by HNF transcription factors. Science 303 , 1378-1381. Ohnishi, S., Yasuda, T., Kitamura, S., and Nagaya, N. (2007). Effect of hypoxia on gene expression of bone marrow-derived mesenchymal stem cells and mononuclear cells. Stem Cells 25 , 1166-1177. Ozer, A., and Bruick, R.K. (2007). Non-heme dioxygenases: cellular sensors and regulators jelly rolled into one? Nat Chem Biol 3, 144-153. Parmar, K., Mauch, P., Vergilio, J.A., Sackstein, R., and Down, J.D. (2007). Distribution of hematopoietic stem cells in the bone marrow according to regional hypoxia. Proc Natl Acad Sci U S A 104 , 5431-5436. Pavesi, G., Mereghetti, P., Zambelli, F., Stefani, M., Mauri, G., and Pesole, G. (2006). MoD Tools: regulatory motif discovery in nucleotide sequences from co- regulated or homologous genes. Nucleic Acids Res 34 , W566-570. Peyssonnaux, C., Nizet, V., and Johnson, R.S. (2008). Role of the hypoxia inducible factors HIF in iron metabolism. Cell Cycle 7, 28-32. Pokholok, D.K., Harbison, C.T., Levine, S., Cole, M., Hannett, N.M., Lee, T.I., Bell, G.W., Walker, K., Rolfe, P.A., Herbolsheimer, E. , et al. (2005). Genome-wide map of nucleosome acetylation and methylation in yeast. Cell 122 , 517-527. Qing, G., and Simon, M.C. (2009). Hypoxia inducible factor-2alpha: a critical mediator of aggressive tumor phenotypes. Curr Opin Genet Dev 19 , 60-66. Rankin, E.B., Biju, M.P., Liu, Q., Unger, T.L., Rha, J., Johnson, R.S., Simon, M.C., Keith, B., and Haase, V.H. (2007). Hypoxia-inducible factor-2 (HIF-2) regulates hepatic erythropoietin in vivo. J Clin Invest 117 , 1068-1077. Rankin, E.B., Rha, J., Selak, M.A., Unger, T.L., Keith, B., Liu, Q., and Haase, V.H. (2009). Hypoxia-inducible factor 2 regulates hepatic lipid metabolism. Mol Cell Biol 29 , 4527-4538. Raval, R.R., Lau, K.W., Tran, M.G., Sowter, H.M., Mandriota, S.J., Li, J.L., Pugh, C.W., Maxwell, P.H., Harris, A.L., and Ratcliffe, P.J. (2005). Contrasting properties of hypoxia-inducible factor 1 (HIF-1) and HIF-2 in von Hippel-Lindau-associated renal cell carcinoma. Mol Cell Biol 25 , 5675-5686. Rhee, S.Y., Wood, V., Dolinski, K., and Draghici, S. (2008). Use and misuse of the gene ontology annotations. Nat Rev Genet 9, 509-515. Richard, D.E., Berra, E., Gothie, E., Roux, D., and Pouyssegur, J. (1999). p42/p44 mitogen-activated protein kinases phosphorylate hypoxia-inducible factor 1alpha (HIF-1alpha) and enhance the transcriptional activity of HIF-1. J Biol Chem 274 , 32631-32637. Rosenfeld, E., Beauvoit, B., Rigoulet, M., and Salmon, J.M. (2002). Non-respiratory oxygen consumption pathways in anaerobically-grown Saccharomyces cerevisiae: evidence and partial characterization. Yeast 19 , 1299-1321.

Page 94 References

Salnikow, K., Aprelikova, O., Ivanov, S., Tackett, S., Kaczmarek, M., Karaczyn, A., Yee, H., Kasprzak, K.S., and Niederhuber, J. (2008). Regulation of hypoxia- inducible genes by ETS1 transcription factor. Carcinogenesis 29 , 1493-1499. Saxonov, S., Berg, P., and Brutlag, D.L. (2006). A genome-wide analysis of CpG dinucleotides in the human genome distinguishes two distinct classes of promoters. Proc Natl Acad Sci U S A 103 , 1412-1417. Schofield, C.J., and Ratcliffe, P.J. (2004). Oxygen sensing by HIF hydroxylases. Nat Rev Mol Cell Biol 5, 343-354. Seglen, P.O. (1976). Preparation of isolated rat liver cells. Methods Cell Biol 13 , 29-83. Semenza, G.L. (2003). Targeting HIF-1 for cancer therapy. Nat Rev Cancer 3, 721- 732. Sharrocks, A.D. (2001). The ETS-domain transcription factor family. Nat Rev Mol Cell Biol 2, 827-837. Shaulian, E., and Karin, M. (2001). AP-1 in cell proliferation and survival. Oncogene 20 , 2390-2400. Shi, Y., Lan, F., Matson, C., Mulligan, P., Whetstine, J.R., Cole, P.A., and Casero, R.A. (2004). Histone demethylation mediated by the nuclear amine oxidase homolog LSD1. Cell 119 , 941-953. Sitkovsky, M., and Lukashev, D. (2005). Regulation of immune cells by local-tissue oxygen tension: HIF1 alpha and adenosine receptors. Nat Rev Immunol 5, 712- 721. Sluimer, J.C., and Daemen, M.J. (2009). Novel concepts in atherogenesis: angiogenesis and hypoxia in atherosclerosis. J Pathol 218 , 7-29. Sutherland, H., and Bickmore, W.A. (2009). Transcription factories: gene expression in unions? Nat Rev Genet 10 , 457-466. Taylor, P.C., and Sivakumar, B. (2005). Hypoxia and angiogenesis in rheumatoid arthritis. Curr Opin Rheumatol 17 , 293-298. To, K.K., Sedelnikova, O.A., Samons, M., Bonner, W.M., and Huang, L.E. (2006). The phosphorylation status of PAS-B distinguishes HIF-1alpha from HIF-2alpha in NBS1 repression. EMBO J 25 , 4784-4794. Tsukada, Y., Fang, J., Erdjument-Bromage, H., Warren, M.E., Borchers, C.H., Tempst, P., and Zhang, Y. (2006). Histone demethylation by a family of JmjC domain- containing proteins. Nature 439 , 811-816. Vaupel, P., Kelleher, D.K., and Hockel, M. (2001). Oxygen status of malignant tumors: pathogenesis of hypoxia and significance for tumor therapy. Semin Oncol 28 , 29-35. Wenger, R.H., Stiehl, D.P., and Camenisch, G. (2005). Integration of oxygen signaling at the consensus HRE. Sci STKE 2005 , re12. Wierstra, I. (2008). Sp1: emerging roles--beyond constitutive activation of TATA-less housekeeping genes. Biochem Biophys Res Commun 372 , 1-13. Wiesener, M.S., Jurgensen, J.S., Rosenberger, C., Scholze, C.K., Horstrup, J.H., Warnecke, C., Mandriota, S., Bechmann, I., Frei, U.A., Pugh, C.W. , et al. (2003). Widespread hypoxia-inducible expression of HIF-2alpha in distinct cell populations of different organs. FASEB J 17 , 271-273. Wouters, B.G., and Koritzinsky, M. (2008). Hypoxia signalling through mTOR and the unfolded protein response in cancer. Nat Rev Cancer 8, 851-864. Xia, X., Lemieux, M.E., Li, W., Carroll, J.S., Brown, M., Liu, X.S., and Kung, A.L. (2009). Integrative analysis of HIF binding and transactivation reveals its role in maintaining histone methylation homeostasis. Proc Natl Acad Sci U S A 106 , 4260-4265.

Page 95 References

Yang, A., Zhu, Z., Kapranov, P., McKeon, F., Church, G.M., Gingeras, T.R., and Struhl, K. (2006). Relationships between p63 binding, DNA sequence, transcription activity, and biological function in human cells. Mol Cell 24 , 593- 602. Yee Koh, M., Spivak-Kroizman, T.R., and Powis, G. (2008). HIF-1 regulation: not so easy come, easy go. Trends Biochem Sci 33 , 526-534. Young, R.M., Wang, S.J., Gordan, J.D., Ji, X., Liebhaber, S.A., and Simon, M.C. (2008). Hypoxia-mediated selective mRNA translation by an internal ribosome entry site-independent mechanism. J Biol Chem 283 , 16309-16319. Zambelli, F., Pesole, G., and Pavesi, G. (2009). Pscan: finding over-represented transcription factor binding site motifs in sequences from co-regulated or co- expressed genes. Nucleic Acids Res 37 , W247-252. Zhang, X., Odom, D.T., Koo, S.H., Conkright, M.D., Canettieri, G., Best, J., Chen, H., Jenner, R., Herbolsheimer, E., Jacobsen, E. , et al. (2005). Genome-wide analysis of cAMP-response element binding protein occupancy, phosphorylation, and target gene activation in human tissues. Proc Natl Acad Sci U S A 102 , 4459-4464. Zheng, Y., Josefowicz, S.Z., Kas, A., Chu, T.T., Gavin, M.A., and Rudensky, A.Y. (2007). Genome-wide analysis of Foxp3 target genes in developing and mature regulatory T cells. Nature 445 , 936-940.

Page 96 Supplements

13. Supplements

13.1. Top 300 up regulated genes in PMH and PMM

Top 300 Genes >1.5 fold upregulated in PMM Top 300 Genes >1.5 fold upregulated in PMH Fold Fold Gene Name p-value Change Gene Name p-value Change

Egln3 0.000100 20.74 Cyp2s1 0.000036 136.00

Ankrd37 0.002185 13.58 1190002H23Rik 0.000077 81.61

Adm 0.000998 11.62 Chdh 0.000780 36.81

Ak3l1 0.002748 11.54 Anxa8 0.000141 35.44

Ndrg1 0.000983 9.08 Dok7 0.000683 33.70

Bnip3 0.005619 7.67 Stmn4 0.003565 27.56

P4ha2 0.001501 6.02 Kbtbd11 0.001028 23.07

Ero1l 0.003712 5.82 Plcxd1 0.001679 21.67

Slc16a3 0.001855 5.70 Sema7a 0.001440 20.62

Agpat9 0.003343 5.02 Iqck 0.000439 18.49

Rcor2 0.015189 4.76 Gprc5b 0.000439 18.49

Gys1 0.001811 4.68 Tgm1 0.002479 17.64

F13a1 0.002372 4.24 Smtnl2 0.002474 17.52

Aldoc 0.000872 4.22 Prr15 0.000617 17.45

Supt6h 0.005340 5.60 Plod2 0.001114 17.27

Prelid2 0.001560 4.04 4930583H14Rik 0.000578 16.17

Pgk1 0.000961 4.31 Adm 0.000295 16.15

Selenbp1 0.009057 3.94 Il12rb1 0.001063 14.36

Slc2a1 0.000894 3.89 Egln3 0.001547 14.25

Tmem45a 0.005783 3.88 Tmcc3 0.001685 13.72

Tpi1 0.001413 3.78 Grhl1 0.001005 13.68

AC161410.3 0.002375 3.77 Amz1 0.000214 13.01

Prr15 0.003668 3.77 Vegfa 0.002778 14.22

Slc7a2 0.000578 3.70 Ankrd37 0.004703 12.23

Rgs11 0.000252 3.68 Ptges 0.000715 11.66

1190002H23Rik 0.009236 3.68 H2-Ab1 0.000138 12.62

Selenbp2 0.009305 3.65 Rcor2 0.000992 11.05

2310056P07Rik 0.000712 3.60 Pkp3 0.000171 10.92

Pfkfb3 0.000108 3.56 Pgf 0.000769 10.18

Pgm2 0.004730 3.45 Cidea 0.000486 10.00

Cxcr7 0.004420 3.40 5730557B15Rik 0.010427 9.97

Hyal1 0.000750 3.49 Vldlr 0.000958 9.62

P4ha1 0.011339 3.37 Ndrg4 0.001505 9.58

Sox7 0.007087 3.15 Pfkfb3 0.001431 9.48

Mif 0.001107 3.09 Pim1 0.008687 9.29

Mthfd1l 0.016038 3.06 CT010467.6 0.012259 9.20

Ldha 0.003130 3.48 BC031353 0.003555 9.20

Page 97 Supplements

AL929165.9 0.004132 3.03 Ndrg1 0.000188 9.12

AL845449.4 0.004132 3.03 Pfkp 0.006706 9.74

AC153577.2 0.004132 3.03 Tmeff1 0.005535 9.01

Serpinb1c 0.012104 3.02 Ccbp2 0.004816 8.95

AL663049.8 0.005295 2.98 Epb4.1l4a 0.000213 8.83

AV249152 0.005295 2.98 Bnip3 0.000146 8.79

AL671335.12 0.002440 2.97 4933402N22Rik 0.009261 8.70

AC163215.4 0.001356 2.96 Car2 0.002617 8.51

Pglyrp3 0.001356 2.96 AC138119.5 0.008692 8.43

Grhpr 0.001607 2.95 AC164410.5 0.007620 8.43

Mid1 0.003817 2.91 Unc13a 0.001381 8.02

Serpinb1a 0.009044 2.85 Slc16a3 0.000209 8.01

Kit 0.011188 2.81 St8sia3 0.001047 8.29

AL845256.3 0.008215 2.80 Nppb 0.004399 7.87

AL672249.6 0.002740 2.79 AC132468.4-202 0.002043 7.75

Rasgef1a 0.009598 2.78 EG432825 0.002043 7.75

2600010E01Rik 0.001525 2.75 AC087117.9-203 0.013922 7.89

Jun 0.007707 2.68 Gpr120 0.000449 7.51

Tmem42 0.005763 2.67 AC136642.4-203 0.002649 7.23

Jmjd1a 0.004449 2.97 Fbxo10 0.000287 7.08

Mamdc2 0.012989 2.66 Polr1e 0.000287 7.08

Pdk1 0.010072 2.65 Adora2b 0.001050 7.07

Kcna7 0.003409 2.60 1700025G04Rik 0.001071 6.92

Tph1 0.007339 2.59 Cldn1 0.002996 6.90

AC160757.3-201 0.003932 2.58 Foxk1 0.001295 6.87

Cav1 0.005121 3.08 Arg1 0.004818 6.86

BX005181.5 0.002848 2.58 Psmd1 0.000895 6.81

Eno1 0.002848 2.58 Htr2b 0.000895 6.81

AC150274.2 0.002848 2.58 AC136642.4-202 0.003007 6.80

Htra3 0.000757 2.56 Sdcbp2 0.008005 6.74

Fzd7 0.000512 2.51 Pdk4 0.006252 6.72

Pkm2 0.005857 2.48 Speer3 0.005640 6.69

Pgam1 0.000612 2.47 Syce2 0.002415 6.68

Zfyve28 0.001480 2.46 1700001L05Rik 0.002801 6.68

Jmjd2b 0.001686 2.45 Krt19 0.006360 6.67

Dgkg 0.005519 2.42 Hk1 0.009403 7.70

Spata13 0.008979 2.37 Lrrc58 0.000747 6.56

Serpine1 0.004829 2.36 Enah 0.000706 6.51

Eif4ebp1 0.004476 2.36 AC141567.4-202 0.006021 6.40

Mxi1 0.003631 2.35 Lce1d 0.003657 6.34

Gpi1 0.000400 2.32 Slc2a1 0.003617 6.31

Tec 0.013569 2.31 Icam2 0.006975 6.28

AL731692.8 0.007518 2.30 Gys1 0.007887 6.20

Mast4 0.009726 2.30 Tbl2 0.003026 6.19

Zfp395 0.015158 2.29 Colec12 0.008947 7.66

Page 98 Supplements

Tcp11l2 0.002484 2.28 Wdfy1 0.007536 6.16

Pafah1b3 0.004994 2.26 Ero1l 0.001062 6.13

Mmp13 0.011935 2.24 Ugcg 0.001655 6.12

Trim13 0.010468 2.20 9130404D14Rik 0.003173 6.08

Pfkl 0.003313 2.19 Klf4 0.003683 6.03

Il15 0.003473 2.19 Ablim1 0.002940 5.90

Arhgap22 0.015659 2.18 Prelid2 0.003602 5.88

Btg1 0.000469 2.17 Fabp4 0.005920 5.88

AC133650.4 0.003186 2.17 Nupr1 0.001645 5.86

Traf6 0.001618 2.16 Gsn 0.007575 5.79

E230019M04Rik 0.003262 2.15 Tmem189 0.000123 5.68

AL672270.12 0.003262 2.15 Lce1f 0.003013 14.64

Syne2 0.009858 2.13 AC127247.3 0.003993 5.68

Foxk2 0.011923 2.12 Flt1 0.015077 5.65

EG277333 0.001558 2.11 1810015C04Rik 0.002336 5.63

EG624367 0.001347 2.11 Atf3 0.002957 5.59

Peli2 0.001347 2.11 Kcne3 0.002047 5.56

AC170187.2 0.001490 2.11 Erg 0.010678 5.53

2310016C08Rik 0.005198 2.11 AL691472.6 0.000383 5.41

AL672180.11 0.002359 2.10 Tspan18 0.000383 5.41

Pcnx 0.001762 2.34 Jarid2 0.001511 5.39

AL805906.7 0.002607 2.10 Arl4d 0.000574 5.38

Cntln 0.002607 2.10 Adamtsl5 0.011697 5.33

Kbtbd11 0.003125 2.09 Krt17 0.002863 5.32

5330426P16Rik 0.001330 2.51 Fzd1 0.001821 5.27

Mtss1 0.003298 2.09 Slc20a1 0.001641 5.22

Il1rl1 0.007920 2.08 Cryab 0.000223 5.20

Appl2 0.001204 2.07 Laptm4b 0.000855 5.13

Stard9 0.015959 2.07 4931428F04Rik 0.006380 12.62

Cdan1 0.015959 2.07 Nol3 0.012542 5.02

Maff 0.019152 2.05 Rragd 0.000610 4.96

EG545052 0.003161 2.05 Dusp4 0.001779 4.96

Higd1a 0.008220 2.22 P4ha2 0.003607 4.93

1190002N15Rik 0.004045 2.04 Lonrf3 0.010236 4.93

AC168279.3 0.000638 2.03 2310047D13Rik 0.000279 4.89

Bckdhb 0.000262 2.03 Dusp9 0.003173 4.87

Spsb4 0.010982 2.02 Eif4a2 0.002359 5.13

OTTMUSG00000003456 0.013907 2.01 Ppfibp2 0.001512 4.87

Nt5e 0.003961 2.01 Gpsm1 0.007336 4.78

Pfkp 0.001488 1.99 P4ha1 0.004661 4.75

Rora 0.012614 2.07 Hmox1 0.003084 4.72

Vegfa 0.003964 2.35 Kdelr3 0.000115 4.66

Stk24 0.009957 1.98 Cugbp2 0.002054 4.65

Col6a3 0.005629 1.97 Fgf1 0.013113 4.60

AC115124.6 0.002253 1.96 Krt23 0.002195 4.55

Page 99 Supplements

AC156551.5 0.002253 1.96 Mapk13 0.001578 4.53

Tmem189 0.007558 1.96 Myd116 0.003441 4.52

3000002C10Rik 0.002170 1.96 Rassf4 0.016532 4.51

A430107O13Rik 0.013185 1.95 8430408G22Rik 0.016532 4.51

Il13ra1 0.003975 1.94 2310040A07Rik 0.007960 4.41

Cox7a1 0.004536 1.94 Col12a1 0.000631 4.41

Samd9l 0.018739 1.94 St3gal1 0.002460 4.37

Vim 0.000895 1.93 Aff4 0.002111 4.35

Trib3 0.001585 1.93 Nampt 0.004369 4.33

AL604043.11 0.000895 1.92 Zfand2a 0.002281 4.32

Jmjd6 0.012132 1.92 Plekha2 0.005762 5.95

AL731648.6 0.000556 1.91 Creb3l3 0.000928 4.24

Asph 0.005434 1.91 Tnfrsf10b 0.000302 4.34

Ccng2 0.001269 1.91 Tiparp 0.002012 4.23

Rlf 0.006235 1.97 Arl5b 0.000687 4.21

2610024E20Rik 0.004993 1.90 Lce1g 0.004045 4.19

Scd2 0.006105 1.90 4632417N05Rik 0.000832 4.17

Ahnak 0.018101 1.89 Lce1i 0.006496 4.16

Sdc4 0.002789 1.89 RP23-256J17.1 0.006666 4.14

Snx25 0.007885 1.88 Rod1 0.001337 4.11

Arrdc3 0.019577 1.88 Bex2 0.007300 4.11

Crebl2 0.005065 1.87 A230050P20Rik 0.002395 4.09

Npepps 0.006056 1.86 Angptl6 0.002395 4.09

Golph3l 0.002904 2.00 Lamb3 0.003981 4.09

AL954636.9 0.001712 1.85 Dnmt3a 0.001629 4.06

AL929407.16 0.001211 1.90 Arid5a 0.001281 4.06

Neurl2 0.000488 1.84 Rab11fip1 0.002660 4.06

AC087117.9-201 0.010452 1.84 EG385328 0.006321 4.04

Hspa1b 0.010452 1.84 Mef2a 0.000163 4.03

AC087117.9-203 0.003961 2.09 Rab15 0.010421 4.02

AC156499.2 0.015068 1.84 Folr2 0.006442 4.02

Glb1 0.015068 1.84 Dcn 0.004189 4.02

Map2k1 0.008357 1.83 Cdh11 0.001474 4.00

Ier3 0.007942 1.83 Kif21b 0.001924 4.00

AL929132.9 0.002834 1.83 Me2 0.003572 4.00

Dppa3 0.019514 1.83 AC116557.30 0.000242 3.99

Ptpro 0.003290 1.81 Selenbp1 0.002516 3.98

Rcbtb2 0.009412 1.84 2610005L07Rik 0.000910 3.97

Lmo2 0.011616 1.81 Srgn 0.003207 3.96

Errfi1 0.001025 1.81 Cxcr7 0.004409 3.96

Aldoa 0.003651 1.93 Wnt9a 0.003457 3.95

Tspan9 0.001550 1.79 Esam 0.009152 3.88

Alkbh5 0.006630 1.79 Flrt3 0.013098 3.88

Kif21b 0.001917 1.79 Macrod2 0.013098 3.88

Hipk3 0.003252 1.78 AL928700.7 0.013098 3.88

Page 100 Supplements

Ppm1b 0.015795 1.78 Trp53i13 0.001796 3.87

Iqck 0.012069 1.78 RP24-388B10.2 0.002295 3.87

Gprc5b 0.012069 1.78 OTTMUSG00000016789 0.002295 3.87

A230051G13Rik 0.004901 1.77 RP24-388B10.4 0.002295 3.87

Wsb1 0.009613 1.77 Xk 0.002813 3.93

Anxa2 0.005736 1.83 OTTMUSG00000016790 0.002295 3.87

BC031353 0.010128 1.77 RP24-388B10.6 0.002295 3.87

4930431B09Rik 0.005091 1.77 RP24-388B10.8 0.002295 3.87

Dapp1 0.006563 1.76 RP24-388B10.10 0.002295 3.87

Cpa3 0.007797 1.76 OTTMUSG00000016779 0.002295 3.87

Zfp503 0.000259 1.75 1700012L04Rik 0.002813 3.93

Sorbs1 0.011197 1.75 RP24-388B10.9 0.002295 3.87

Pgp 0.000932 1.75 Pfkl 0.002021 3.87

Hk2 0.011760 1.75 Mtss1 0.003469 3.84

Tmem71 0.001182 1.82 1700027L20Rik 0.001138 3.82

E130012A19Rik 0.001915 1.74 Jmjd1a 0.009279 4.02

Slc37a4 0.008871 1.74 Tsc1 0.001516 3.72

Samd10 0.016554 1.74 5430407P10Rik 0.000243 3.72

Hk1 0.002591 1.73 Higd1a 0.000715 3.72

Prnp 0.000048 1.73 Inhbb 0.011098 3.71

Unc13b 0.001955 1.73 Prdx4 0.000231 3.70

Bhlhb2 0.001694 1.72 AI413582 0.000815 3.70

Tagln 0.000542 1.72 AC123048.4 0.000518 3.97

CT030181.13 0.002139 1.72 AC152164.15 0.000518 3.97

AL627074.11 0.008353 1.71 Grpel2 0.004134 3.69

Prelid1 0.008576 1.71 9130227C08Rik 0.000409 3.66

Fbxl15 0.002580 1.71 Lpin2 0.007459 3.65

Csnk1d 0.006571 1.83 Atp13a4 0.000227 3.64

AL662901.18 0.006628 1.71 Timp3 0.001583 3.64

Lancl1 0.003391 1.70 Syn3 0.001583 3.64

Slc38a2 0.008158 1.70 Crybb3 0.002848 3.63

Fbxo21 0.002222 1.70 Cd68 0.001851 3.59

Pde4b 0.014739 1.69 Stk17b 0.007264 3.56

Dsel 0.006893 1.69 Fhdc1 0.004671 3.55

Lamp2 0.001983 1.69 Hspa4 0.006617 3.53

Zfp292 0.000444 1.69 2310016C08Rik 0.004683 3.57

Mll5 0.002080 1.68 Smox 0.009979 3.51

Tmem159 0.004957 1.68 2310008H04Rik 0.007934 3.51

Arntl 0.003473 1.68 Rybp 0.008200 3.50

Rnf113a1 0.001955 1.68 AC192334.1-201 0.008200 3.50

Myo1d 0.008041 1.68 Fgd6 0.004245 3.49

Ppp1r14c 0.014465 1.68 Odf2 0.002109 3.47

Lrrk2 0.012514 1.67 Alkbh5 0.014473 3.47

Cma1 0.013874 1.67 Selenbp2 0.002733 3.46

Nampt 0.011947 1.67 Uchl1 0.004002 3.46

Page 101 Supplements

Casp6 0.002202 1.67 Appl2 0.003909 3.45

AL671990.5 0.001073 1.67 Rpl22 0.001675 3.44

Olfr1153 0.000551 1.66 Gm22 0.000327 3.44

Hyal3 0.007373 1.65 Dcps 0.003240 3.43

Nat6 0.007373 1.65 4930581F22Rik 0.003240 3.43

Cysltr1 0.010446 1.65 1700113O17Rik 0.014687 3.43

Galnt7 0.005874 1.88 Egln1 0.002466 3.43

Bbc3 0.010635 1.65 Rnf12 0.019883 3.43

Txnip 0.009473 1.65 Dedd2 0.003120 3.42

OTTMUSG00000021867 0.009473 1.65 Serpine1 0.001827 3.41

2310022B05Rik 0.004306 1.65 Camk2d 0.003365 5.45

Psme3 0.003825 1.65 Arrdc4 0.008732 3.38

Cd14 0.000237 1.64 Smurf1 0.006538 3.37

AL808132.5 0.001015 1.64 Apln 0.015273 3.58

AC150660.4 0.001015 1.64 Cd72 0.015602 3.37

AL627070.16 0.001015 1.64 Lcp1 0.001361 3.34

AC142167.4 0.001015 1.64 Ankrd23 0.002915 3.31

AL935328.14 0.001015 1.64 Dgkh 0.007591 3.31

AC156283.6 0.001015 1.64 Krt16 0.002117 3.30

AC124113.9 0.001015 1.64 Samd8 0.001048 3.30

AL669829.11 0.001015 1.64 Map3k1 0.001098 3.29

BX005163.13 0.001015 1.64 Ak3l1 0.007155 3.29

Gapdh 0.001634 1.96 Gpr137b 0.001824 3.28

AC150744.2 0.001015 1.64 Ing5 0.000241 3.28

AC118474.10 0.001015 1.64 Btg2 0.009577 3.32

AL805956.22 0.001015 1.64 Tnfrsf1b 0.004937 3.26

AC163335.6 0.001015 1.64 Apoa4 0.002135 3.25

AC147142.2-201 0.001015 1.64 Pdk1 0.004947 3.25

AL954370.3 0.001015 1.64 Mir16 0.001383 3.25

OTTMUSG00000005300 0.001015 1.64 Sap30 0.006207 3.23

AC134337.3 0.001015 1.64 Mxi1 0.002749 3.22

AC125070.4 0.001015 1.64 Cd109 0.002915 3.20

AL845308.11 0.001015 1.64 Lilrb3 0.014610 3.20

AC121279.7 0.001015 1.64 Adamts1 0.002815 3.96

AL807395.8 0.001015 1.64 Snx8 0.003064 3.20

AC166075.2 0.000862 1.90 Mt1 0.012268 3.55

AC148327.3-203 0.001015 1.64 Ugp2 0.002507 3.17

AC102196.7 0.001015 1.64 Sox9 0.006937 3.16

AL732526.8 0.001015 1.64 Rbm35b 0.001967 3.16

OTTMUSG00000017911 0.001015 1.64 Polr3g 0.009804 3.14

AC134918.5 0.001015 1.64 Klhl24 0.004343 3.12

AC125407.4 0.001015 1.64 Pcgf5 0.003896 3.12

AL671988.15 0.001015 1.64 Fcna 0.000252 3.12

BX679665.3 0.001015 1.64 OTTMUSG00000012511 0.000252 3.12

CT009534.18 0.001015 1.64 Aldoa 0.010994 3.09

Page 102 Supplements

AC121959.3 0.001015 1.64 Stx3 0.003397 3.09

AL807790.15 0.001015 1.64 Adarb1 0.002488 3.06

AC166827.2-201 0.001015 1.64 AC140331.2-201 0.009936 3.06

AC107864.11 0.001015 1.64 Tram2 0.002412 3.05

AL772328.13 0.001015 1.64 Wdr33 0.002226 3.70

AL607064.16 0.001015 1.64 Zfp655 0.006051 3.05

AC158396.2-202 0.001015 1.64 1190002N15Rik 0.000845 3.04

AC132147.4 0.001015 1.64 Zxdc 0.000162 3.03

1700027N10Rik 0.001787 1.64 Ptpn14 0.002864 3.03

Casp4 0.002021 1.64 2310021P13Rik 0.011891 3.02

Zfp7 0.010802 1.64 Mt2 0.010777 3.01

Adipor2 0.012738 1.63 Bnip3l 0.005236 3.01

Spn 0.000793 1.63 CT030242.6 0.009613 3.01

Pde4a 0.015422 1.63 Ormdl3 0.007867 3.01

Gata2 0.005682 1.63 Oxct1 0.008969 3.08

Cdkn1a 0.012854 1.63 1500032D16Rik 0.000676 2.99

Map3k6 0.005080 1.63 Ralgds 0.013324 2.98

Narf 0.007097 2.12 AC159809.2-201 0.004543 2.98

Cdkn1b 0.001577 1.62 Ccng2 0.001966 2.97

AC122193.5 0.001577 1.62 Centd3 0.004149 2.97

Cd3eap 0.014366 1.62 Rnf144b 0.002449 2.96

D12Ertd553e 0.005049 1.62 Myo5a 0.000140 2.95

Rest 0.003287 1.85 Eif1b 0.015867 2.95

Cep170 0.002740 1.62 Slc46a3 0.004098 2.94

Cav2 0.002684 1.62 RP24-302M3.2 0.000945 2.94

Gadd45a 0.009031 1.62 Gadd45a 0.003676 2.94

Polr3g 0.010965 1.61 Arrb1 0.014854 2.94

Cadm1 0.004915 1.61 Phf3 0.004922 2.93

Klhl6 0.016141 1.61 Mysm1 0.001263 2.93

Tle1 0.003876 1.71 2310056P07Rik 0.000516 2.93

Frmd6 0.001563 1.60 Hook2 0.005067 2.93

Pnrc1 0.001282 1.60 Akap2 0.002921 2.92

Man1a 0.002474 1.60 Mapk7 0.011942 2.91

Heca 0.000048 1.59 Inhba 0.009922 2.91

Rchy1 0.006711 1.59 Fabp1 0.011055 2.90

Trappc6a 0.008213 1.59 Zfp654 0.016651 2.90

AL929226.7 0.000620 1.59 Xirp2 0.014524 2.90

Prdx4 0.000115 1.59 Pkm2 0.002890 2.90

Page 103 Supplements

13.2. Group III genes of PMM, PMH and Raw.264 cells (Top 300)

Group III genes in PMM Group III genes in PMH Group III genes in Raw.264 (Top 300)

Gene Name p-value Distance Gene Name p-value Distance Gene Name Reads Distance

2310016C08Rik 4.89E-05 -97 1110058L19Rik 0.000156867 597 Pdk1 740 -24

2310044G17Rik 2.86E-05 -189 1500031H01Rik 7.31E-05 1626 Atxn7l2 676 -63

2310056P07Rik 1.27E-06 -247 1810063B05Rik 0.000519906 2188 Hjurp 624 58

4930583H14Rik 4.59E-07 -105 2010012O05Rik 0.000583294 130 Eno1 610 -356

5830415F09Rik 1.06E-05 -29 2310016C08Rik 2.15E-05 -692 Ankrd37 608 -463

6030408C04Rik 0.000422619 -42 2310056P07Rik 2.01E-07 -426 Gbe1 568 359

9030624J02Rik 0.000596294 -4668 2610510H03Rik 0.000254009 1743 D030013I16Rik 564 -390

Abca4 0.000444898 240 2810405K02Rik 0.000343609 1552 Gipr 543 -93

Ablim1 0.000265013 -3876 2810422J05Rik 0.00040683 580 Plod2 525 -169

AC133650.4 0.000505192 -3 2810428I15Rik 0.00040683 -3048 Narf 514 -526

AC139884.8 0.00013712 -106 4833442J19Rik 3.65E-05 -7 Gapdh 510 -166

Adamts20 0.000447399 -1433 4930519F16Rik 0.000263597 -3935 Efna1 497 -203

Agpat5 0.000417623 280 4930583H14Rik 0.000152161 -204 Ier3 495 -260

AI317395 0.000368012 1563 4932418E24Rik 0.000465378 -4030 Kcnab2 493 121

Alg11 0.000351884 0 6030408C04Rik 4.33E-05 20 Sfi1 489 -4221

Alkbh5 3.9E-08 -609 6430517E21Rik 0.000329079 -2823 Seh1l 467 -90

Ankrd23 0.000435355 -807 8430408G22Rik 0.000159231 -41 Gpi1 454 -3935

Ankrd37 8.13E-10 -478 A930104D05Rik 0.000279238 -447 Pgam1 443 -269

Ankzf1 6.21E-05 -24 AC091531.9 3.81E-05 163 Asph 441 -158

Anxa2 4.74E-06 -116 AC107671.7 3.2E-05 -1533 Bsg 435 -115

Arg1 2.17E-06 -2898 AC165946.4 0.000566094 -235 Pkm2 420 558

Asb1 0.000470373 -4509 Acox1 0.000152246 -275 2310016C08Rik 411 -1439

Atp2b3 5.9E-05 1307 Adam22 0.000161433 1678 Pgm2 398 317

Atpbd4 1.39E-06 -17 Adm 0.000459532 -880 Rsbn1 396 -321

B3galnt1 0.000291925 -746 Adm2 0.000130401 605 Ccdc58 393 -445

Bat5 0.000281291 -3675 Aff3 0.000543129 -1194 2310056P07Rik 393 -118

Bhlhb2 0.000326481 -211 Agpat4 0.00018956 153 Pnrc1 392 853

Birc3 0.000108541 3 Alkbh5 2.54E-06 -1022 AC116557.30 389 -1149

Blm 0.000222553 -148 Ankrd1 0.000447919 -3529 Pkp2 387 -189

Bnip3 1.77E-07 -411 Ankrd37 6.93E-06 -711 Gys1 376 -342

Bnip3l 0.000106903 169 Ankzf1 0.000430117 -206 Ruvbl2 376 -250

Btg2 0.000360406 -4307 Anp32a 0.000450142 1017 Map3k1 375 -763

C1qb 0.000464954 -602 Anxa2 6.65E-05 -264 Arrb1 375 873

Car8 0.000563025 390 Aoc2 0.000420004 -4439 Jmjd6 373 562

Ccdc126 0.000210067 -146 Ap3b2 0.000352235 -810 Fzd7 369 -494

Ccdc58 1.27E-06 -316 Ascl1 5.09E-05 2349 CT009708.6 368 -2483

Page 104 Supplements

Ccng2 0.000256376 -271 Astn1 0.000329079 -3271 6030408C04Rik 367 -191

Cenpl 0.000518533 -83 Atf7ip 0.000434322 584 Rnf19a 365 -156

Chaf1a 0.000234755 -3638 Atp12a 0.000562777 1633 Park7 364 -4096

Chrm1 0.000273685 -4363 Atp1a2 0.000435616 2144 1600014C23Rik 362 -2958

Clca3 3.81E-05 -2471 Atpbd4 7.4E-05 -188 Zfp292 349 -23

Clcnkb 0.000210258 1690 Atr 3.81E-05 67 4930583H14Rik 345 -258

Cnga1 0.000508985 2401 B230317F23Rik 0.000454662 -124 Pfkl 343 370

Cope 0.00039855 -216 B3galt5 0.000289985 -3930 Bhlhb2 339 -201

Csdc2 0.000211307 1889 Bbs5 0.000539171 -4799 Rnmt 334 -179

CT009708.6 4.74E-06 -2545 Bhlhb2 5.89E-05 -297 4933403F05Rik 334 -14

Ctsa 4.6E-06 -338 Bnc2 0.000130663 -901 Tpi1 328 547

D130059P03Rik 1.19E-05 -276 Bnip3 1.64E-06 2090 Neud4 326 -139

Dars2 0.000518533 -52 Bnip3l 0.000546102 478 Ero1l 319 -319

Ddit4 8.45E-06 -35 Btn2a2 0.000598022 -1873 Neurl2 317 25

Ddx49 0.00039855 -125 C130057D23Rik 0.000584781 -3811 Ctsa 317 530

Dhcr24 0.000216807 -2846 Capn6 9.48E-05 598 Phospho1 316 -196

Disc1 0.000555922 -4145 Ccdc58 2.01E-07 -137 AC120398.10 315 -23

Dnajc5 0.000364843 -177 Cd200r1 9.92E-05 -1675 P4ha2 314 -124

E230015B07Rik 0.000224817 2996 Cdk2ap2 0.000410758 2362 9130227C08Rik 314 -210

Ehf 0.000448706 -396 Cdx2 0.000106753 -3319 Cbln3 314 -503

Eif4ebp1 1.66E-05 747 Chrnb3 0.000199967 695 Hmga1 308 -3731

Eno1 1.72E-09 -252 Clcn3 0.000454662 -13 Bnip3 308 -120

Ero1l 2.99E-06 -294 Cplx4 0.00023064 -460 Ldha 307 -70

F10 4.45E-06 -375 Cpxm2 0.000312883 1538 B230317F23Rik 306 23

F3 0.000265988 -1661 Csmd1 0.000160835 -1148 Jarid1b 305 -3804

Fbln2 0.000241778 -868 CT009708.6 6.65E-05 -2397 Car12 304 2

Fcamr 0.000363412 209 Ctnnal1 2E-05 -3567 Mrpl18 302 -325

Fgfrl1 0.000373886 2829 Cyfip2 0.000162875 -2271 Tcp1 302 -49

Fkbp15 0.000323032 -37 Cyp26a1 5.83E-06 -2193 Nos2 298 -211

Foxg1 0.000236391 -3765 D10Wsu102e 0.00023258 -562 Alkbh5 294 -843

Fzd7 1.45E-06 -418 D16Ertd472e 0.000467519 333 Mettl11a 293 -159

Gapdh 1.29E-09 -196 D830014E11Rik 0.000309408 -4403 Gabpb2 293 -64

Gata1 0.000570003 -2982 D930020E02Rik 0.000471988 62 Eif4ebp1 292 730

Gbe1 5.47E-05 525 Dars 0.000110062 361 Adm 291 2897

Gmppa 0.000221067 -29 Ddit4 1.08E-06 -324 Pfkp 291 1485

Grid1 0.000402777 515 Ddx3y 0.000434088 -4657 Slc16a3 290.6666667 -22

Grin2a 0.000318681 -1260 Dedd 0.000132478 -1166 2310008H04Rik 289 96

Gys1 8.52E-09 -107 Def8 0.000135857 -1991 Eno2 288 -5

Gzf1 0.000256361 -104 Defb36 0.000463888 -2110 Rbpj 287 -3225

Hils1 0.000359148 973 Dnaja2 0.000110259 -2309 Fosl2 287 -2061

Hk2 2.75E-05 -245 Dync1h1 0.000591415 -3857 Me2 285 -203

Hsp90ab1 1.32E-05 -860 Dzip1l 0.000191363 -3606 Dennd4b 285 -734

Ier3 1.99E-06 -101 Eda2r 0.000372587 -4723 Stc1 285 366

Inhbb 0.000174645 -807 Ehbp1 0.000268342 -4812 Wsb1 280 -85

Isca1 0.000270342 -1827 Emr4 2.49E-05 1552 Rnf126 278 -1

Page 105 Supplements

Jarid1b 0.000167022 -3773 Eno1 5.09E-08 -83 Gtf2e2 275 -108

Jarid1c 2.84E-05 -52 Enox1 0.000464633 1811 1700104B16Rik 275 -1122

Jmjd1a 1.74E-09 -637 Enpep 0.000108422 -562 Bhlhb3 271.5 -808

Jmjd2b 3.76E-07 -454 Epm2a 0.000204509 1825 Anxa2 270 -178

Jmjd6 5.11E-06 313 Epyc 0.00036961 -4711 Nampt 269 -264

Krt13 0.000441895 -2581 Errfi1 0.000346055 -61 Bnip3l 268.5 -148

Krt15 0.00010803 -1805 Fads2 0.000305306 -3191 Safb2 267 416

Lace1 0.000442202 -2207 Fcer1a 0.000145694 -3992 Selenbp1 266 -170

Ldha 1.88E-05 -149 Fcrla 9.48E-05 1048 Triobp 266 -3289

Lgals3 0.000227677 -4827 Fdx1 9.5E-05 2100 Atp11b 264 -264

Lpar4 0.000327807 -1020 Ferd3l 0.000355863 -404 Mif 259 -43

Ly6g6f 0.000281291 -45 Fgf3 0.000217148 1329 Hyal1 258 624

Mettl9 4.21E-07 -10 Fhl2 0.000513895 -3934 Nat6 258 -656

Mrpl45 5.5E-05 -239 Fkbp14 0.000135422 -1674 Ticam2 252 85

Mrpl54 7.28E-05 30 Fntb 0.000408548 3 Rcor2 251 -71

Mthfd1l 0.000323064 577 Folr4 1.32E-05 1168 F10 247 -672

Myt1 0.000324867 -1309 Foxp3 2.13E-05 -471 F3 245 -1736

Nampt 0.000555601 -187 Fyb 0.000468949 2085 1700112E06Rik 245 -410

Narf 5.54E-07 -166 Gabrb3 3.15E-06 1685 Arid2 245 -787

Ncln 0.000339303 -534 Gapdh 2.87E-07 -341 Srgn 243 -2965

Neurl2 4.6E-06 -3 Gas6 0.000361564 -1122 Hk2 242 -216

Nobox 0.000234398 905 Gc 0.00034019 -2800 Narg1l 241 -182

Obfc2a 0.000198978 127 Ggh 0.00014476 2356 Pcgf5 240 -244

Otog 0.00018548 -2250 Gjc1 0.000566796 679 Tnfrsf9 236 -2502

Oxsr1 0.0001923 -615 Gpr1 5.12E-05 -2963 Mel13 236 -21

P2ry4 0.000190337 -4599 Gramd3 0.000532832 -1556 Higd1a 234 -134

P4ha1 3.1E-06 30 Gtf2f1 3.45E-05 -372 5830415F09Rik 233 -44

P4hb 1.16E-05 -552 Gys1 3.64E-06 -178 Prelid2 231 -172

Pcdha10 0.000543374 -3126 Hdhd1a 0.000369122 -3076 Sap30 230 -742

Pcm1 0.000381192 -136 Hint2 0.000493374 1274 Rnf7 230 -10

Pcsk9 0.000384107 1142 Hivep1 0.000554647 -1465 Cep170 228 650

Pde1c 2.99E-05 2472 Hoxd9 0.000547429 -2058 BC030867 227 -3540

Pfkfb3 2.42E-06 -1157 Ict1 0.000591056 -4782 Fgf11 227 609

Pfkl 1.13E-05 144 Isca1 0.00025836 -2177 Rusc2 226 2267

Pgd 5.17E-05 -62 Iscu 6.05E-05 -889 Dnajc5 225 -41

Pgk1 0.00013486 3 Isg20 1.12E-05 -306 Rabggta 224 -2453

Pgm2 0.000409109 242 Isyna1 5.63E-05 -1128 6330569M22Rik 223 -165

Piga 0.000380997 -311 Iyd 0.000257197 463 Pfkfb3 220.6666667 -2442

Pkm2 4.63E-09 503 Jmjd1a 5.78E-07 -552 Klhl35 218 2835

Pkp3 0.000301651 905 Jmjd2b 0.000207 -208 2900016B01Rik 217 907

Pnrc1 7.32E-05 641 Kcnh7 0.00012726 -2505 Ccng2 217 -268

Polr2d 0.000285688 56 Kctd6 0.000222827 1087 Mthfd1l 217 582

Ppp2r2b 0.000460329 1748 Khk 9.04E-05 -456 Isca1 216 -1725

Prelid1 4.66E-05 123 Kif11 0.000175903 -66 Rlf 213 8

Prr15 0.000479882 1177 Klf7 0.000517891 -715 Clca5 212 -3228

Page 106 Supplements

Pygl 5.17E-06 -70 Klhl1 0.000588366 -1672 Prr12 212 -968

Rasl10b 0.000447551 -2120 Klhl9 0.000185321 216 Stt3b 212 -515

Rbm3 9.32E-05 116 Krt84 0.000346879 -542 Ahnak 212 -4463

Rgma 0.000486255 -816 Lactb2 1.77E-05 -121 P4hb 211.5 -579

Rnf145 8.18E-06 999 Ldha 8.19E-05 -1187 l7Rn6 210 -148

Rnf152 0.000538257 -3189 Lmln 5.27E-05 -1719 Amz1 210 -131

Rps6kl1 0.000470054 -1299 Lrp1 0.000228891 -1335 P4ha1 209 -110

Rusc2 1.37E-05 2221 Map2k7 2.69E-05 -99 Gmppa 209 -8

Ruvbl2 8.52E-09 -485 Map3k1 4.91E-06 -1005 Jmjd1a 209 -526

S1pr4 0.000339303 2740 March1 0.000509414 -3608 Map4k4 203 933

Safb2 0.000481919 377 Mcl1 0.000411434 68 Nktr 202 -205

Sall3 0.000291988 -2215 Mrpl45 6.14E-05 -239 Lonp1 201 -25

Sf3b1 4.54E-06 -206 Mrps28 0.000274517 37 Galk1 199 347

Sfrs1 0.000524429 -130 Ms4a13 0.000292547 -492 Jarid2 199 -675

Sfrs11 0.00022596 -1694 Mtcp1 0.000508149 -4302 Lgals3 198 544

Sfxn5 0.000204439 1081 Narf 3.34E-06 -166 Pygl 196 -100

Sh3gl1 0.000234755 -201 Ndfip1 0.000313729 819 Helb 195 -136

Slc16a3 1.18E-07 1990 Ndrg4 0.000383702 522 Cd47 195 -171

Slc26a5 0.000473083 -2202 Noxo1 0.000399622 99 Tas2r119 194 -2473

Slc31a1 0.000323032 -135 Npat 0.000336362 1744 Arhgdig 193 -1576

Smtnl2 0.000472642 -272 Nrxn3 6.59E-05 1272 Zfp335 193 -582

Snca 5.39E-05 -315 Obfc2a 4.74E-05 127 EG624866 193 -406

Sntb1 0.000581609 -4881 Ogt 6.76E-05 -130 Rgs11 193 -13

Spcs3 0.000186706 -132 Olfr1121 0.000435625 -2797 Slc2a1 192 -2871

Spo11 0.000137378 -474 Olfr1140 0.000374575 1003 Hdac7a 192 -2006

Srebf1 0.000150093 86 Olfr1261 0.000290036 -325 Stc2 192 -1313

Stx18 0.000270225 -13 Olfr134 0.000484949 -913 Pim3 190 -623

Sytl2 0.000474651 -151 Olfr1419 0.000593748 -2773 2810008M24Rik 190 -2851

Tcap 0.000591625 -4582 Olfr1469 5.34E-05 -2356 AC116591.4-202 188 13

Tcfap4 1.69E-05 900 Olfr1471 0.000421478 -3750 Dock8 187 -81

Tfrc 5.4E-05 43 Olfr173 0.000464473 -1766 Pgk1 186 -186

Thbs2 0.000409987 -1835 Olfr178 0.000438918 -1311 Aldoc 186 -124

Tmem112 3.15E-05 72 Olfr340 0.000281199 -92 Bbs5 185 52

Tmem194 1.14E-07 616 Olfr494 0.000194191 -1175 Tcfap4 182 1118

Tmem42 0.000505192 -79 Olfr622 0.00013418 -827 Kif21b 182 -4395

Tnfaip3 0.000399453 -2909 Olfr640 8.05E-05 1108 Sf3b1 182 16

Tpi1 8.94E-08 660 Olfr731 0.000404882 -774 Ddit4 181 -171

Trappc2 0.000179996 -1380 Olfr788 0.000318105 -281 Atr 180 -159

Trappc6a 8.27E-07 88 Olfr794 0.000496575 -2834 Epm2a 179 -43

Trim29 0.00042545 1936 Olfr851 0.000335301 1942 Myom3 179 1514

Ttll3 0.000320873 -4644 Olfr961 0.000595807 1623 Sertad1 178 -469

Ube2g1 2.57E-06 1399 Oxsr1 9.97E-05 -615 Foxo3 178 -51

Usf2 5.5E-07 -1007 P4ha1 1.64E-05 -141 Vwa1 177 -112

Ush1c 0.00018548 -247 Pak3 0.000557225 -1848 1700029J07Rik 176 -53

Vdac1 0.000274455 -236 Pax1 0.000380726 -269 Tgif1 176 -39

Page 107 Supplements

Vgll4 8.58E-05 1202 Pcyt1b 0.000448898 486 Tusc3 176 469

Wdr69 0.000154401 1303 Pdk1 0.000208794 -36 Ufsp2 176 -228

Zbtb33 0.000348424 -3751 Pfkfb3 1.56E-05 -1288 Colec12 175 1083

Zfp280c 0.000454488 -365 Pfkl 0.000140087 43 Pcm1 175 -229

Zfp62 8.14E-05 309 Pgd 0.000156156 -211 Micall2 175 -3921

Pgk1 0.000211405 -39 Mt1 174 -197

Pgm2 0.0003883 242 Agrp 173 1473

Phka1 0.000140741 -1570 Atp6v0d1 173 -816

Pitpnm1 0.000410758 -337 C030039L03Rik 172 -2961

Pkm2 5.17E-05 734 AC160757.3-201 172 203

Pkp3 0.000579347 527 Setd5 172 -1142

Plch1 8.92E-05 581 Igf2bp2 172 2315

Plekha8 0.000135422 -260 R3hdm1 171 644

Pnrc1 0.000111904 771 Glt1d1 171 658

Pole 0.000260098 2853 Orai2 171 -604

Ppp1r12a 0.00012602 1269 Spesp1 171 -2292

Ppp1r3f 2.13E-05 -4957 Vdac1 170 -504

Praf2 0.00034833 -846 Egln1 169 -247

Prelid1 0.00031609 183 Bcl7b 168 -15

Prok2 0.000588081 1752 Fblim1 168 -55

Prox1 0.00028482 -2091 Lgmn 167 1596

Prph2 0.000128936 383 Acvrl1 166 472

Pth2r 0.000382335 -1675 Pvr 166 -59

Ptma 0.000405631 -98 Plekha2 166 34

Puf60 0.000164542 -134 Eef2k 165 -79

Qk 0.000531373 1115 Ube2g1 164 1231

Rabac1 0.000430796 -1395 Sap18 163 -140

Rabggta 1.49E-05 -2696 Fusip1 160 -115

Rbpj 0.000103609 -1102 Rab39 159 118

Rel 0.000531225 315 Ube2q1 159 -283

Rnf145 6.24E-05 860 4632404H12Rik 159 -93

Rps11 0.000303783 -68 2510039O18Rik 159 -1998

Rps6ka6 1.87E-05 -4945 Plod1 159 -2172

Ruvbl2 3.64E-06 -414 Aldoa 159 -136

Sall1 2.22E-05 962 Artn 157 2710

Sart3 6.05E-05 -265 Bnip1 157 -144

Sele 0.000312375 -1212 Tuba1b 156 348

Serpina10 0.000521693 2805 Tnfsf9 155 24

Serpinb11 0.000562937 -1091 EG665044 154 255

Serpine1 1.55E-05 -226 Fosb 154 1316

Sf3b1 9.28E-05 178 Klf13 154 -843

Shf 0.000107294 639 Hspb7 154 -2109

Slc16a3 2.78E-06 1990 Braf 154 -175

Slc2a1 6.78E-07 -2523 Clcnkb 154 -4078

Slc36a4 0.000449738 -2584 Tpd52 153 -76

Page 108 Supplements

Slc39a12 0.000562912 -325 Mettl9 153 -26

Slc41a3 0.00026551 -893 Cry2 152 2900

Smad2 0.000278453 1425 Gbf1 152 810

Snap25 0.000311153 16 Fes 151 -102

Snapc1 0.000570455 811 AI848100 151 -201

Snx30 0.000428566 1857 Serpine1 151 -182

Spata19 0.000505916 682 Pgp 151 -509

Spats1 0.000565605 1169 Cdkn1a 150 -2850

Spink6 0.00034762 1332 Dynll1 150 -208

Stx6 0.000107616 -2478 Ttc14 149 -145

Sync 0.00056425 732 Suv420h1 148 589

Tacr3 0.000459486 880 Dap 145 -210

Taf13 0.000257534 -1619 Exoc1 145 -4544

Tbc1d25 6.11E-05 760 Erp29 145 -34

Tcf4 0.000192441 -20 Zfp446 145 -91

Tessp2 0.000442568 673 AC168063.3 145 -3382

Tfrc 0.000469868 43 Tlr6 145 -1164

Tifa 3.9E-05 -2910 C79267 145 -1283

Tiprl 0.000458386 -72 Map2k1 144 442

Tmem147 0.000436318 -299 Lrba 144 -580

Tmsb4x 0.000348767 -861 4930543L23Rik 143 -45

Tnfrsf19 0.000173121 1448 Spg21 143 276

Tnrc18 0.000235095 1188 Tbc1d4 143 226

Tpi1 0.000374024 660 A230051G13Rik 143 380

Trib3 3.55E-05 -93 Fbxl11 142 -4

Trpc7 0.000547418 -1948 Haao 141 -4149

Tsga14 0.000370265 -3595 Tbkbp1 141 -531

U2af2 0.000523532 -521 1110020G09Rik 140 108

Ube2g1 0.000475695 1604 Llgl1 140 -19

Ubp1 0.00037749 1788 Dars 140 498

Usf2 0.000135198 -890 Adssl1 140 -193

Usp54 0.000474489 -209 Utp11l 140 -179

Wdr23 5.28E-05 521 Cd3eap 139 -1734

Wdr40a 0.000140001 -22 AC139884.8 139 -248

Wwp2 0.000367019 -141 AC142098.2 139 -2521

Zdhhc16 0.000513009 2497 Ing3 139 288

Zfp384 3.37E-05 -3804 C130050O18Rik 139 -2391

Zscan18 0.000553094 2061 Ppp1r13l 139 -63

Dixdc1 138 -260

Ccdc126 138 -127

Hook2 137 -4655

Usp28 137 -309

Hddc2 137 266

Tnfaip3 137 -2811

Jmjd2b 136 -510

Page 109 Supplements

Agl 136 -48

Raver1 136 247

Exoc7 135 -149

Blm 135 -104

Slc41a2 135 -37

Ubl7 135 -119

Maz 134 32

BC031781 134 -589

Ttc25 134 2710

Oxsr1 133 -491

5330426P16Rik 133 446

4933426M11Rik 133 -2634

Mrpl45 133 -249

Mxi1 133 2581

Hk1 132.5 -202

Alg11 132 19

Agpat5 132 331

Ankrd24 131 -1001

Sirt6 131 -77

Rapgef1 131 1578

Prkcbp1 131 -4478

Ralb 131 1864

1110039B18Rik 131 -89

Gzf1 131 -162

Cstf2 131 -82

AC087780.10 130 181

Gpr68 130 82

Ppp1cb 130 -127

Snx21 129 -111

Zfp652 129 -510

Lemd2 129 1257

Tmem112 128 -7

Stk33 128 -4900

Atf3 128 -2616

Ache 127 -1105

Sh3gl1 127 -222

Usf2 127 -893

Gls 127 -273

Kdelr2 127 -19

Page 110