myc mediated transcriptional modulation sanjay j....

MYC MEDIATED TRANSCRIPTIONAL MODULATION

SANJAY J. CHANDRIANI

A DISSERTATION PRESENTED TO THE FACULTY OF PRINCETON UNIVERSITY

IN CANDIDACY FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

RECOMMENDED FOR ACCEPTANCE BY THE DEPARTMENT OF MOLECULAR BIOLOGY

SEPTEMBER 2005

© Copyright by Sanjay J. Chandriani, 2005. All rights reserved.

ii

ABSTRACT

Deregulated Myc expression can lead to profound physiological changes in cells,

including increased cell proliferation, oncogenic transformation, apoptosis and inhibition of differentiation. Myc is a transcription factor that can both stimulate and inhibit transcription and is believed to affect the physiological changes by coordinate regulation of many Myc responsive genes.

To define the global transcriptional response to Myc, we undertook an expression microarray analysis of deregulated Myc expression in primary human fibroblast cells. In this analysis, we identified a core Myc gene expression signature that could be used to analyze the expression profiling datasets of human cancers. With this approach, we were able to identify malignant tissue samples that display the Myc gene expression profile. For example, basal-like breast carcinomas often displayed the Myc signature.

One hallmark of transformed cells in vitro is a reduced dependence on serum for cell survival and proliferation. To determine if deregulated Myc expression underlies this reduced dependence on serum by regulating serum response genes, we were able to interrogate our data for those genes that had previously been defined in the core serum response (CSR). Many CSR genes responded to both Myc and serum similarly. However, the complete CSR was not recapitulated by Myc overexpression.

Using a genetic approach to identify proteins that can complement the slow growth phenotype in cells lacking Myc expression, we were unable to identify a single gene product (other than Myc family members) that could completely complement the phenotype. One gene product, mitochondrial serine hydroxymethyltransferase (mSHMT), was identified in the genetic screen to partially complement the growth phenotype and was subsequently characterized as a direct Myc target gene. This enzyme catalyzes reactions that provide a major source of the one carbon unit that is required for many metabolic processes.

Myc can stimulate the expression of a gene that is silent in most somatic cells. This gene, hTERT, is the catalytic subunit of telomerase and has been found to be expressed in most malignant cells. We demonstrated a required role for the recruitment of TRRAP and the associated histone acetyltransferase activity for Myc mediated transactivation of the hTERT gene.

iii

ACKNOWLEDGEMENTS

I am lucky to have the help and support of so many people. In some small way, I hope this page captures my gratitude. I thank Mike Cole for many things. First and foremost, for taking me into his lab and allowing me unfettered access to his often exciting and unconventional ideas. Furthermore, I am truly grateful for the scientific space he gave me to make mistakes and to learn how to think independently without fearing disparagement with failed attempts. As evidence of the leeway I was permitted, chapter two of this thesis is exactly that. With limited informatics skills in the lab, Mike let me begin an expensive and risky project that has deeply influenced how I can and will likely do science in the future. Through the ups and downs of this project, Mike always challenged me to bring the discussion of p-values, correlation scores, overlapping responses back to nuts and bolts biology. His interpretations and reluctance to easily accept bioinformaticians’ positions have forced me to truly understand and critically consider the new scientific world views.

Jane Flint has become like a second advisor in the last two years. I can’t thank her enough for her kindness in taking me into the lab when I couldn’t move to New Hampshire with Mike. I found her door to always be open for discussions of issues ranging from science to personal. Jane has been so deeply engaged in and genuinely caring about my work and life outside the lab that I know I will feel a bit rudderless as I leave Princeton. Jane has always demanded excellence everywhere ranging from how an experiment is performed and interpreted to how a sentence is written. Importantly, however, she has taught me that a no nonsense work ethic goes well with a softer side that recognizes the frailties of human nature.

Mike Whitfield was my first true mentor in the world of bioinformatics. From teaching me how to plan microarray experiments to actually performing and interpreting them, I am forever grateful. Mike has provided excellent guidance and critique of each step of the analyses in the second chapter. I could never have learned so much so quickly without his openness and willingness to help.

I thank my other committee members, Jim Broach and Paul Schedl, in their guidance and encouragement throughout my years at Princeton. I thank David Botstein for early guidance on the project described in chapter 2. The impression of a lab experience is often colored by the people who are working alongside in the trenches, (or benches). I have been so lucky to be in the same bay as my classmate, Daniel Miller, because of the camaraderie, the support, the demand for scientific excellence, the wrestling/punching matches in the early years, the coffee breaks in the later years, the discussions both about and not about science and most importantly for the friendship that will be lifelong. My friendship with Guangli Wang has enriched my life. Through hours of debates regarding the sins and virtues of capitalist America and communist China, we have come to intimately understand and appreciate each other. I will long remember and appreciate my friendship with Misha Nikiforov that has developed from long discussions about science and politics. We have worked productively together on several projects and I am grateful for his guidance and scientific exchange in these endeavors. Iulia Kotenko supported the whole lab with her contagious high spirits and energy. I am

iv

grateful for her interest and caring in my pursuits. I thank Victoria Cowling for being a very supportive and engaged friend in the lab.

Before I began my final project in the laboratory, when someone said, “perl,” I thought, “jewelry.” I have been on a perl/matlab/unix learning curve in the past year and two amazing people have made this curve steep: John Matese and Chad Myers. They both have gone far beyond normal courtesy of helping a desperate newbie for which I am deeply grateful.

In the years that I have worked in the department, I have had the privilege of working with so many smart and interesting people in both the Cole and Flint labs. Collectively, they have helped make my graduate training enjoyable. The Zakian lab members have been a rich source for technical experience of which I took full advantage. I am thankful for their openness and patience with my repeated advice-seeking excursions.

In collaboration with Chuck Perou at UNC-Chapel Hill, we were given access to a breast cancer profiling dataset that had not been published. This dataset was particularly useful in our studies because it had been generated using the same microarray platform we used for our studies. Eirik Frengen who was a visiting scientist in the Perou lab and now has his own lab at the University of Oslo was instrumental in giving clinical relevance to the Myc signature in fibroblast. Eirik, in fact, made a key suggestion on how to define the Myc signature. I thank Eirik for his responsiveness, advice and patience. Tina DeCoste provided excellent training and guidance when my projects required the use of the fluorescence activated cell sorting. Her advice dramatically lowered the activation energy to begin a new experimental approach. Norma Caputo knew how to efficiently deal with the administrative jungle of paperwork when it came to reimbursements, special orders and grant applications. She was always willing to help in these tasks even for us lowly graduate students. As she retires next month, I wish her a fond farewell. The graduate school experience couldn’t have been balanced and complete without having rich and meaningful relationships outside the laboratory. Our friendships with three families in particular, (Dan, Claire & Sarene), (Ritin, Jasmeet & Karan) and (Brian, Jennifer and Taylor) were the salve for difficulties, the source of many joys, the well of encouragement and the hope for tomorrow. I will always cherish my time with our friends and hope our paths will cross many more times. I am grateful for being warmly welcomed into my new family through my marriage. They have all been patient and supportive of my work in spite of my experimental difficulties that were also shouldered by their daughter and sister. My good friend, Radha-Vallabha, has been an astounding and non-judgmental listener to whom I have consistently turned for personal guidance. He continues to encourage and support my journey towards truth. Mom and Dad have showered me with boundless love and encouragement. They have stoked my passion for understanding how things work by letting me take apart radios with no real hope of them ever being functionally put back together, by giving me my own “workshop” where my friend and I built a go-cart that wouldn’t fit through the workshop door, by demanding an excellence in my work by setting examples in their own. For these reasons and so many more, I am lucky and grateful to have my parents as

v

my parents. Kinnari has been a paragon of faith in me. However, through her perseverance and diligence in her own endeavors, she is actually a role model for me.

Mira is my raison d’être. She brings smiles without trying. She brings light in the darkest of hours. Although I am sure she is going to regret it at times, I am so excited and grateful to be her father.

I can not list all the obvious and subtle ways that my best friend supports me because there is likely a page limit on a dissertation. I will, however, try to convey my deepest gratitude for her love and faith in me. Graduate school would never have been possible for me without Mimi by my side. Late nights, poor time estimates to complete lab work, miserable company after miserable results, overzealous company after exciting results were always met with patience, encouragement and, most importantly, faith that I could succeed. I don’t deserve her, but I am not going to let her know that. I am eternally grateful for her being in my life.

vi

TABLE OF CONTENTS

ABSTRACT...................................................................................................................... iii

ACKNOWLEDGEMENTS ............................................................................................ iv

TABLE OF CONTENTS ............................................................................................... vii

TABLE OF FIGURES................................................................................................... viii

TABLE OF SUPPLEMENTAL FILES.......................................................................... x

CHAPTER 1: Introduction.............................................................................................. 1 Myc Background ....................................................................................................................... 2 Myc gene expression signature................................................................................................. 6 Complementation Screen .......................................................................................................... 7 TRRAP Dependency ................................................................................................................. 8 Methodological background ................................................................................................... 10

CHAPTER 2: The Myc Induced Gene Expression Signature.................................... 14 Results....................................................................................................................................... 15 Discussion ................................................................................................................................. 72 Materials and Methods ........................................................................................................... 81

CHAPTER 3: A Functional Screen for Myc-responsive Genes Reveals SHMT: a Major Source of One-carbon Unit for Cell Metabolism ............................................. 86

Results....................................................................................................................................... 87 Discussion ................................................................................................................................. 96 Materials and Methods ......................................................................................................... 102 Acknowledgements ................................................................................................................ 105

CHAPTER 4: TRRAP-dependent and TRRAP-independent Transcriptional Activation by Myc Family Oncoproteins.................................................................... 106

Results..................................................................................................................................... 107 Discussion ............................................................................................................................... 127 Materials and Methods ......................................................................................................... 133 Acknowledgements ................................................................................................................ 136

REFERENCES.............................................................................................................. 138

vii

TABLE OF FIGURES

CHAPTER 2:

Figure 2.1 Analysis of expression from transgenes………………………….. 17

Figure 2.2 Tamoxifen treatment……………………………………………… 19

Figure 2.3 The p53 response to MycER activation………………………….. 21

Figure 2.4 RNA quality assessment…………………………………………... 21

Figure 2.5 RT-PCR analysis of cad, cyclinD2 and hTERT…………………. 23

Figure 2.6 Validation of normalized microarray data……………………… 26

Figure 2.7 Figure of Merit analysis of time course data……………………. 29

Figure 2.8 k-means clustering using 15 clusters of time course data……… 30

Figure 2.9 Hierarchical clustering of time course and steady state data….. 32

Figure 2.10 Overrepresented GO terms – time course experiments………. 33

Figure 2.11 Overrepresented GO terms – steady state experiments………. 35

Figure 2.12 Serum responsiveness…………………………………………… 40

Figure 2.13 Venn diagram: Myc responsive genes in different studies……. 41

Figure 2.14 The Myc gene expression signature…………………………….. 43

Figure 2.15 E-box overrepresentation……………………………………….. 46

Figure 2.16 The Myc gene expression signature in breast tumor tissue ….. 49

Figure 2.17 Cluster and survival analysis…………………………………… 51

Figure 2.18 Quantification of the Myc signature in breast tumor tissue….. 54

Figure 2.19 c-Myc protein in normal and basal breast carcinoma tissue…. 56

Figure 2.20 The Myc gene expression signature in prostate tissues……….. 58

Figure 2.21 The Myc gene expression signature in normal human tissues... 61

viii

Figure 2.22 The Myc signature vs. c-Myc levels in normal human tissues.. 63

Figure 2.23 The Myc gene expression signature in B-cell lymphomas……... 65

Figure 2.24 Serum stimulation vs. Myc overexpression…………………….. 69

Figure 2.25 Myc signature in the serum response………………………….... 71

CHAPTER 3: Figure 3.1 Outline of the experimental strategy……………………………… 88

Figure 3.2 Overexpression of mSHMT enhances proliferation of c-myc-null cells…………………………………………………………. 91

Figure 3.3 mSHMT and cSHMT are Myc-responsive genes………………... 93

Figure 3.4 c-Myc binds to the promoters of mSHMT and cSHMT genes in vivo…………………………………………………………….. 97

CHAPTER 4: Figure 4.1 Activation of endogenous genes by c-Myc, c-Myc∆ and L-myc… 110

Figure 4.2 Recruitment of nuclear cofactors by L-Myc and N-Myc………... 112

Figure 4.3 E1A-N-Myc∆ chimeras selectively precipitate TRRAP from 293 cells…………………………………………………………… 114

Figure 4.4 E1A-N-Myc∆ chimeras restore the transactivation potential of N-Myc∆……………………………………………………. 117

Figure 4.5 Distinct phenotypes of c-Myc, L-Myc, N-Myc and E1A-N-Myc∆ chimeras in cell growth and primary cell transformation…………………………………………………………. 120

Figure 4.6 Recruitment of histone acetyltransferase activity by Myc family proteins………………………………………………... 123

ix

TABLE OF SUPPLEMENTAL FILES

All supplemental files referenced in this thesis are available at the following web address:

http://genomics-pubs.princeton.edu/chandriani_dissertation_2005/

Table 2.1: GO term finder results for time course experiment (sheet 1)

GO term finder results for steady state experiment (sheet 2)

Table 2.2: Myc cancer database target genes converted to unigene.

Table 2.3: Clustered normal human tissues array labels.

PDF:

Nikiforov MA, Chandriani S, Petrenko O, O’Connell B, Kotenko I, Beavis A, Sedivy JM, Cole MD 2002. Functional screening for Myc-induced genes reveal serine hydroxymethyltransferase, a major source of one-carbon unit for cell metabolism. Molecular and Cellular Biology 22: 5793-5800.

PDF:

Nikiforov MA*, Chandriani S*, Park J*, Kotenko I, Matheos D, Johnsson A, McMahon SB, Cole MD 2002. TRRAP-dependent and TRRAP-independent transcriptional activation by Myc family proteins. Molecular and Cellular Biology 22: 5054-5063. *Authors equally contributed.

x

CHAPTER 1: Introduction

1

Myc Background

The MYC proto-oncogene is an essential gene whose function is required for the

normal development of organisms ranging from mice to simpler metazoans such as fruit

flies (Charron et al. 1992; Davis et al. 1993; Gallant et al. 1996). From gene disruption

experiments of MYC family members, it has become increasingly evident that Myc plays

an important role in the control of normal cell proliferation and growth. Disruption of N-

MYC or c-MYC in mice leads to embryonic lethality. Conditional gene disruption in

mouse embryonic fibroblast cells was characterized by a marked reduction in cell

proliferation (de Alboran et al. 2001). In Drosophila, a hypomorphic allele of dMYC

(diminutive) gives rise to flies with small body size and female sterility (Gallant et al.

1996). Somatic cell knockout of c-MYC in immortalized rat fibroblasts causes reduced

protein synthesis rates and a lengthening of the cell doubling time by almost three fold.

While MYC function has been extensively studied in normal contexts, attention

was first drawn to MYC in the context of oncogenic transformation by an avian tumor

retrovirus (MC29) (Sheiness and Bishop 1979; Vennstrom et al. 1982). The cellular

homolog of the transforming oncogene of MC29 was determined to be c-MYC

(Vennstrom et al. 1982; Colby et al. 1983; Watt et al. 1983). Soon after these seminal

discoveries, other mechanisms that resulted in abnormally regulated MYC expression

were identified, including retroviral insertions proximal to the cellular proto-oncogene,

chromosomal rearrangements and gene amplifications (review, (Facchini and Penn

1998)). Some of these deregulatory events can cause MYC to function as a potent

oncogene and it is estimated that 20% of all human cancers harbor an oncogenic allele of

MYC (Nesbit et al. 1999).

2

To help elucidate how Myc affects normal organismal development and

oncogenic transformation, enormous strides have been made in identifying the

biochemical properties of the MYC family gene products (herein referred to as Myc).

For example, Myc has been shown to be a transcription factor that can both activate and

repress target genes (review, (Dang 2005)). In fact, a continually growing list of Myc

responsive genes forms the core of a publicly available Myc database

(www.myccancergene.org) (Zeller et al. 2003). Upon dimerization with Max via their

helix-loop-helix/leucine zipper domains, Myc can bind DNA in a sequence specific

fashion (Blackwood and Eisenman 1991). The canonical Myc/Max binding site is

CACGTG, however closely related variants of this sequence such as CACATG,

CATGTG and CACGCG can also be bound by Myc/Max (Blackwell et al. 1993). Other

functionally important regions of Myc have been mapped. Small deletions of

evolutionarily conserved domains (called Myc boxes I and II; MBI, MBII) disrupt some

of Myc’s known biological functions (Stone et al. 1987; Freytag et al. 1990; Evan et al.

1992).

Several Myc-associated proteins have been shown to be required for Myc induced

transformation. TRRAP (transformation/transactivation domain associated protein) is a

large protein (~400kDa) that has been shown to bind Myc (McMahon et al. 1998). Anti-

sense mediated downregulation of TRRAP expression inhibits Myc’s ability to transform

cells (McMahon et al. 1998). Similarly, Tip49 and Tip48, two other Myc associated

proteins which can heterodimerize, have been shown to be critical for Myc mediated

transformation by expression of a dominant negative form of Tip49 in transformation

assays (Wood et al. 2000). Interestingly, recruitment of these cofactors by Myc occurs

3

http://www.myccancergene.org/

via interactions with the MBII domain. Recent biochemical studies of transcriptional

coactivator complexes have identified, among other proteins, TRRAP, Tip49 and Tip48

as components of these complexes. For example, the human counterpart of the yeast

SAGA complex contains TRRAP, the human homolog of GCN5 or PCAF, but not

TIP49/TIP48 (Ogryzko et al. 1998). Another TRRAP containing complex which is

regarded as the human form the NuA4 complex has the TIP60 H2/H4 HAT as well as

TIP49/TIP48 (Ikura et al. 2000). The discovery of TRRAP and TIP49/TIP48 as cofactors

essential for Myc-induced transformation provided an important link between chromatin

modification and Myc-dependent phenotypes (McMahon et al. 1998; McMahon et al.

2000; Wood et al. 2000). As TRRAP can draw the HAT (histone acetyltransferase)

activity of GCN5 or/and Tip60 to in vitro substrates, Myc may modulate transcription by

focusing these activities to specific target gene promoters. The Tip48 and Tip49

ATPase/helicase proteins associated with Myc may prove to be directly involved in Myc

mediated transcriptional regulation.

Exhaustive studies have recognized Myc’s diverse biological activities.

Depending on the context, Myc can increase cellular proliferation, transformation,

apoptosis, genetic instability and also inhibit cellular differentiation (reviews, (Dang et al.

1999; Pelengaris et al. 2002; Soucek and Evan 2002)). Some experiments identifying

Myc’s role in normal cellular proliferation are described above. Transformation of NIH

3T3 cells by Myc was the original assay used to describe the oncogenic properties of v-

and c-Myc (Cooper 1982; Keath et al. 1984). Subsequently, Myc, in cooperation with

Ras (G12V), was shown to transform primary rodent fibroblast cells (Land et al. 1983).

Mice carrying the inducible transgene, MycER, are susceptible to de novo tumorigenesis

4

upon MycER activation (Felsher and Bishop 1999b; Pelengaris et al. 1999). Given

Myc’s role in cell proliferation and transformation, it is counterintuitive that Myc can

induce apoptosis and cell cycle arrest in some contexts. High expression levels in most

cultured mammalian cells coupled with growth factor or serum withdrawal results in

widespread apoptosis (Askew et al. 1991; Evan et al. 1992). Similarly, activation of

MycER in primary human cells activates a G2 cell cycle checkpoint that is accompanied

by double stranded DNA breaks and aneuploidy (Felsher and Bishop 1999a). Indeed,

Myc mediated tumorigenesis is often accompanied by other genetic events that are likely

to mitigate these growth inhibitory activities of Myc. As an example of Myc’s ability to

impede differentiation, when Myc was constitutively expressed in mouse

erythroleukaemia cells, the DMSO-inducible differentiation program was blocked

(Coppola and Cole 1986).

Despite these advances in identifying the biochemical properties and biological

activities of Myc, our understanding of precisely how Myc’s biochemical activities

culminate in these biological activities of Myc remains incomplete. Certainly Myc’s

primary biochemical function is to regulate transcription. Therefore the identity of the

genes which are regulated by Myc should provide insights into how Myc affects various

biological functions. To date, many Myc responsive genes have been identified.

However, few have been identified as being directly regulated by Myc

(http://www.myccancergene.org/site/mycTargetDB.asp). In addition to experiments that

identify transcriptional responsiveness to Myc, the most compelling evidence supporting

a direct role for Myc in transcriptional regulation is the ability of Myc to bind regulatory

sites of target genes in vivo. Such bona fide Myc target genes have gene products which

5

are involved in cell proliferation (Cdk4, Cylin D2, Id2, Cul1, etc), nucleotide

biosynthesis (CAD) and polyamine biosynthesis (ODC) (Bello-Fernandez et al. 1993;

Wagner et al. 1993; Miltenberger et al. 1995; Lasorella et al. 1996; Hermeking et al. 2000;

O'Hagan et al. 2000). Despite a growing number of Myc-responsive genes, the activities

of Myc, especially the control of cell proliferation, have not been attributed to the

transcriptional regulation of a particular gene or group of genes.

Myc gene expression signature New and old technologies are beginning to permit a better appreciation of the

complexity of the cellular response to Myc. In fact, numerous high-throughput studies

querying expression changes induced by Myc such as differential display, serial analysis

of gene expression (SAGE), and expression microarray analysis have been published

(review, (Dang 2005)). Similarly, chromatin immunoprecipitation followed by genomic

microarray analysis (ChIP on chip) and a novel Myc-methylase chimeric protein to

chemically mark Myc binding sites have been used to identify chromosomal locations of

Myc (Fernandez et al. 2003; Orian et al. 2003; Cawley et al. 2004). While many studies

have focused on the discovery of Myc target genes, few have examined the complete

cellular transcriptional changes induced by Myc. In each study, some of the Myc

responsive genes are known to participate in the biological functions affected by Myc.

Indeed, identified Myc responsive genes include those whose gene products are involved

in ribosome biogenesis, protein synthesis, cell cycle regulation and cell adhesion (review,

(Dang 2005)). Strangely, however, the several studies have not identified the same set of

Myc responsive genes. Several differing experimental variables may explain the

apparent incongruity which is discussed later.

6

In chapter 2, we report an attempt to define a fundamental Myc gene expression

profile that is less subject to many of the common experimental variables using an

expression microarray technology. Upon identification of the gene expression signature

based on two different cell culture models of Myc overexpression, we used this profile to

direct query of gene expression data from human cancers. This approach allows us to

isolate the specific transcriptional response to a unique perturbation in a genetically

tractable and easily controlled system and look for this response in datasets from clinical

samples in which many unknown genetic and environmental perturbations have likely

taken place. Similarly, other datasets generated by making different unique

environmental or genetic perturbation in vitro can be used to compare and contrast

responses. Within this body of work, I have contributed the bulk of the analysis. Eirik

Frengen, a visiting scientist in the Perou lab, performed the survival and cluster analyses.

Michael Cole ran one set of microarrays (the steady state arrays) using RNA made by

Mark Mellon and Victoria Cowling.

Complementation Screen Another approach to identifying mechanisms of Myc action is to search for genes

that can complement the Myc loss of function phenotype. In chapter 2, we describe a

genetic screen to isolate those genes encoding products that can complement the slow

growth phenotype displayed by an immortalized rat fibroblast cell line upon genetic

disruption of both c-MYC alleles. The work in this chapter is presented as a modified

version of a manuscript published in Molecular & Cellular Biology (Nikiforov et al.

2002a). Using a novel complementation screen, we identified one gene, mitochondrial

serine hydroxymethyltransferase (mSHMT), that could partially decrease the doubling

7

time of the Myc-null cells. The gene products of mSHMT and its cytoplasmic isoform,

cSHMT, are nuclear genes encoding enzymes in the one carbon unit metabolism pathway.

SHMT enzymes catalyze the conversion of serine and tetrahydrofolate (THF) into

glycine and 5,10-methylene-THF, respectively. This reaction is a major source of one-

carbon-substituted folate cofactors necessary for a variety of metabolic reactions,

including thymidylate and de novo purine biosynthesis, biosynthesis of methionine and

histidine, and the catabolism of methyl groups and formate (review, (Snell et al. 2000)). It

has also been suggested that SHMT activity may regulate cell growth and proliferation

(37, 38, 41) (Thorndike et al. 1979; Snell 1980, 1985). Subsequent experiments

identified both mSHMT and cSHMT as Myc responsive genes. Within this body of work,

I was able to show experimentally that Myc is bound to the promoters of mSHMT and

cSHMT which provides strong support for the conclusion that both mSHMT and cSHMT

are direct Myc target genes.

TRRAP Dependency We were interested in determining mechanisms of activation of specific cellular

promoters in the context of their native chromosomal location. We focused on four

promoters, CAD, CDK4, HSP60, and TERT, that have previously described binding sites

for Myc/Max heterodimers (Miltenberger et al. 1995; Boyd and Farnham 1997;

Hermeking et al. 2000; Frank et al. 2001). However, TERT regulation differs

dramatically from the other genes in exponentially growing cells, since TERT is silent in

almost all somatic cells (Greider 1999), whereas the other genes are ubiquitously

expressed. Interestingly, Myc is the only transcription factor capable of interacting with

the TERT promoter and inducing its expression from the chromosomal locus (Wang et al.

8

1998). Proteins like E2F or E1A, which in many in vivo assays demonstrate activities

similar to that of Myc, fail to activate TERT (Wang et al. 1998). Since TERT expression

is up-regulated in the vast majority of human cancer cells, it was of particular interest to

address how this normally silenced gene can be activated by Myc.

The role of histone acetylation in transcriptional activation by Myc is not clear.

On the one hand, a number of studies reported that activation of the Myc target genes

CAD and TERT occurs without concomitant increases in histone H3 or H4 acetylation

(Eberhardy et al. 2000). On the other hand, Myc recruits HAT activity (McMahon et al.

2000), and the recruitment of TRRAP to the promoter of several Myc-responsive genes

following serum stimulation has been associated with induction of H4 but not H3

acetylation (Frank et al. 2001).

We found that the activation of a silent TERT gene in exponentially growing

primary human fibroblasts requires TRRAP recruitment and is accompanied by both H3

and H4 acetylation. We also demonstrate that Myc mutants that lose their ability to

interact with TRRAP fail to activate TERT in primary cells but are still able to activate

transcription of basally expressed genes in both primary cells and myc-null fibroblasts

(Mateyak et al. 1997). An analysis of wildtype L-Myc protein and N-Myc mutants

demonstrated that TRRAP binding was largely dispensable for normal cell proliferation

and the induction of basally expressed genes. The work in this chapter is presented as a

modified version of a manuscript published in Molecular & Cellular Biology (Nikiforov

et al. 2002b). Within this body of work, I was able to experimentally show that Myc

recruits HAT activity to the TERT promoter and that this activity is dependehnt on Myc’s

ability to bind TRRAP.

9

Methodological background Some of the approaches and tools that we used in the work presented in this thesis

have more recently been described. Therefore, I have provided some background

information on these methodological strategies.

Expression microarrays: Expression microarrays provide a powerful platform to

identify global, or near-global, transcriptional responses (review, (Brown and Botstein

1999)). Two-color microarray technology is based on competitive hybridization of two

differently labeled RNA or DNA samples to elements consisting of 60- to 70-mer

oligonucleotides or short DNAs crosslinked to glass slides (Wang et al. 2003). The ratio

of the label-specific fluorescence signals from an individual element provides a measure

of the relative abundance of a given transcript or DNA segment in the original samples.

In the single color Affymetrix microarray platform, one labeled sample is hybridized to

elements consisting of 25-mer oligonucleotides crosslinked to glass slides (Lockhart et al.

1996). Each element consisting of oligonucleotides with a perfect complimentary

sequence to a target transcript has partner element consisting of oligonucleotides that

have one predicted mismatch. The difference in the fluorescence signals from the paired

spots is reported. We believe the two color microarray platform outperforms the single

color for two major reasons. First, because the ratio between two samples is determined

for a given element in the two-color system, non-uniform printing, hybridization or

washing conditions are more robustly normalized. Second, the single color microarray

platform demands that, under the hybridization conditions, a given transcript has

substantially different abilities to hybridize to the perfect match element and the

mismatch element. In principle, elements can be designed to do this. However, in

practice, we expect that many transcripts may hybridize similarly to a 25-mer

10

oligonucleotide that is a perfect match and one that is different by one central nucleotide.

This may lead to under-reporting of truly expressed transcripts. We have therefore used

Agilent’s human 1A version 2 arrays to study transcriptional responses to Myc

overexpression. These arrays have approximately 20,000 elements which map to 15,463

unique unigene clusters (based on the March 2005 unigene build).

GO Term Finder: Because high-throughput approaches yield data that are not

easily interpreted on a gene-by-gene level, many computational algorithms have been

developed that integrate this type of data into a more approachable format. The Gene

Ontology (GO) consortium has developed a controlled vocabulary that allows continual

description of the biological processes, cellular components and biochemical functions of

individual proteins (Ashburner et al. 2000). The GO consortium has also undertaken the

colossal task of annotating studied genes with this vocabulary. Because there is a

systematic annotation, the annotation can be digitized which permits statistical analyses

of groups of genes. The publicly available software, GO Term Finder, allows one to

identify an overrepresented annotation, or GO term, within a user-defined subset of genes

compared to the background representation level (Boyle et al. 2004). In other words,

GOTermFinder allows one to look at a set of similarly responding genes to a given

stimulation and ask whether there is an underlying biological theme common to the

queried genes.

Chromatin IP: Chromatin immunoprecipitation (ChIP) is a technique that allows

one to evaluate interactions between specific proteins and specific DNA regions in vivo

(review, (Kuo and Allis 1999)). Direct and very closely associated protein-DNA

interactions are preserved by a chemical crosslinking step before cells are harvested.

11

Harvested material is subject to sonication that causes DNA to shear to shorter length

fragments without disrupting the crosslinked protein-DNA structures. This material is

probed by immunoprecipitation for a protein of interest. The enriched protein-DNA

complexes are then heated to reverse the chemical crosslinking which permits PCR based

detection of specific DNA segments. This approach was first described to identify

protein-DNA interactions in yeast; we, and others, have modified this technique for

application in mammalian cell systems (Braunstein et al. 1993; Alberts et al. 1998). As

expected, this approach allows one to identify the chromosomal location of sequence

specific DNA binding proteins, however, acetylation state specific antibodies against

histone proteins have permitted us to monitor the histone acetylation state of very specific

genomic regions.

Centroid-Pearson’s correlation: Upon identification of a global expression

response to a given stimulation using expression microarrays, one may want to search for

this response in different expression profiling datasets. To quantitatively describe the

“presence” of a given response in other expression profiles, one can calculate Pearson’s

correlations between the stereotyped expression profile (or, expression centroid) and the

profiles of interest (Chang et al. 2005). The higher the correlation score between two

profiles, the more the two samples resemble each other. A negative correlation score

indicates an opposite response in the two samples.

CSFE: Carboxyfluorescein diacetate succinilmidyl ester (CFSE) is a fluorescein

molecule with a succinimidyl ester functional group and two acetate moieties. CSFE can

freely diffuse into cells across the plasma membrane. Intracellular esterases cleave the

acetate group to yield a fluorescent dye which is now membrane-impermeable. The

12

fluorescent dye does not affect cellular function. After cells have been pulse stained

with CSFE, the quantity of dye, and hence the total CSFE-attributed fluorescence,

decreases by approximately half with each cell division. Because the amount of

fluorescence is directly correlated to the number of rounds of cell division a stained cell

has undergone, this reagent can be used to identify and sort cells growing at different

rates within a population by using fluorescence activated cell sorter (FACS).

13

CHAPTER 2: The Myc Induced Gene Expression Signature

14

Results

Experimental design

The pathology that results from MYC overexpression is undoubtedly the

consequence of altered expression of both direct and indirect target genes. Hence, we

designed our experiments to capture the complete cellular transcriptional response to

Myc overexpression rather than try to limit our study to direct Myc target genes.

Experiments to identify Myc responsive genes were performed with human primary

foreskin fibroblast harboring inducible or constitutive Myc constructs.

N-Myc is a Myc family member that acts in a similar fashion to c-Myc in most

experimental systems (Malynn et al. 2000). We therefore chose to examine N-Myc as a

model member of the Myc family of proteins. A conserved region among Myc family

members, termed Myc-box II (MBII), has been shown to be essential for many of Myc

functions, such as the regulation of some target genes, transformation and apoptosis. In

our study, we performed parallel experiments with the wildtype and MBII deletion

constructs.

Two models of Myc overexpression were employed in this study. In the time

course models, inducible Myc related proteins were activated and cells were collected in

time course fashion. In the steady state model, exogenous MYC genes were

constitutively expressed for an extended period after which the cells are harvested. The

inducible Myc construct we used is pLPC-N-MycERTM. N-MycERTM is a fusion protein

between N-Myc and a portion of the estrogen receptor (Littlewood et al. 1995). This

fusion protein is believed to be sequestered in the cellular cytoplasm until cells are treated

with 4-hydroxytamoxifen (OHT) which results in N-MycERTM translocation to the

15

nucleus where it can regulate the transcription of Myc target genes. The ER portion of

the fusion protein is believed not to influence Myc’s ability to regulate transcription

because of a critical mutation in ER that has destroyed its innate ability to stimulate

transcription (Littlewood et al. 1995). Expression of this construct is driven by a CMV

promoter. The constitutive Myc construct we used is pL(N-Myc)SH. N-Myc expression

of this construct is regulated by the Moloney Murine Leukemia Virus (MoMuLV) long

terminal repeat (LTR) regions in the construct.

The exogenous inducible and constitutive constructs were introduced into primary

human foreskin fibroblast cells by retroviral transduction. Upon viral transduction and

selection of the target cells, cells were expanded and harvested for RNA (in the case of

the steady state constructs) or treated with OHT for increasing time periods (in the case

of the inducible constructs) and then harvested for RNA. Technical replicates for vector,

N-Myc, and N-Myc(∆MBII) cells were labeled for the steady state study. In the

inducible MycER study, untreated samples from the three lines (vector, N-MycERTM, N-

Myc(∆MBII) ERTM) were collected in triplicate, while single samples were collected at

specific times from 15’ to 48 HRS following OHT treatment.

Preliminary experimental steps and system validation

Expression from the exogenous transgenes was confirmed by western blot or RT-

PCR (Figure 2.1). Expression level of wildtype and mutant transgenes were were similar

within each experiment class.

Inducing MycER proteins in primary human fibroblast can cause the activation of

a G2 cell cycle checkpoint (Felsher et al. 2000). This effect can be measured by FACS

analysis. However, the reduction in overall cell proliferation and morphological changes

16

1 2 31 2 3 1 2 3B.A.

1 2 31 2 3 1 2 31 2 3B.A.

Figure 2.1 Analysis of expression from transgenesA. Whole cell lysates of primary human fibroblast cells infected with pLPCX (lane 1) pLPC-N-MycERTM (lane 2), and pLPC-N-Myc(∆MBII)ERTM (lane 3) were resolved by PAGE and transferred to PVDF for western blot analysis. Because exogenous proteins were FLAG-tagged, western analysis was performed using antibodies specificto the FLAG epitope. B. Because the expression level of transgenes driven by the viral LTR is low, RT-PCR was used to monitor transgene expression from cells infected with pLXSH (lane 1) pL(N-Myc)SH (lane 2), pL(N-Myc∆MBII)SH (lane 3). The RNAs used for this analysis were obtained from cells made at a different time than those used for the microarray analysis.

17

are also evident by light microscopic examination of populations of cells either treated

with OHT or not for 48 HRS (Figure 2.2). These pictures are representative of other

plates and other areas of the same plate. Control cells (LPCX) treated with OHT do not

show any morphological changes or growth defects upon OHT treatment. Cells

expressing MycER display considerable morphological and cell number changes upon

OHT treatment. Fewer MycER cells are present after OHT treatment (even though equal

numbers of cells were plated in the treated and untreated plates) and the treated cells are

more round and appear to make fewer contacts with the plate surface. Some

morphological variation is evident in the different cell types even in the absence of OHT

treatment. We believe these changes to be a result of some OHT-independent nuclear

localization of ER fusion proteins.

The arrest in the G2 phase of cycle has been shown to be mediated by p53

(Felsher et al. 2000). Because the MycER expression construct used in these previous

studies was different than the construct we used by several important criteria (the Myc

family member used, which type of ER fusion used, and the retroviral vector used were

all different), we wanted to determine if p53 protein levels were activated in our

experimental system. Furthermore, we wanted to compare the p53 protein level changes

to the previously well characterized p53 response to UV stimulation (Hall et al. 1993).

As shown in figure 2.3, p53 levels were undetectable in control cells treated with OHT,

while p53 levels substantially rise upon OHT treatment in N-MycER cells. Induction of

N-Myc(∆MBII)ER does not lead to a substantial increase in p53 levels, although there

does appear to be higher background levels of p53 in both untreated and treated N-

Myc(∆MBII)ER cells. Similarly, higher background levels are seen in uninduced N-

18

-OHT, 48hrs +OHT, 48hrs

Empty vector

N-MycER

N-Myc-(∆MBII)ER

Figure 2.2 Tamoxifen treatmentHuman primary foreskin fibroblast cells transduced with retrovirus made with pLPCX, pLPC-N-MycERTM, pLPC-N-Myc(∆MBII)ERTM were passed to two plates. The next day, cells were either treated with 0.8 uM 4-hydroxy tamoxifen (+OHT) or not (-OHT). After 48hrs, cells were visualized by light microscopy. These images are representative images of different areas within a tissue culture plate.

19

MycER cells. To compare this increase in p53 levels to previously described p53

mediated responses, we treated cells to increasing doses of UV irradiation. UV irradiated

cells were harvested at two timepoints following UV treatment. The increase in p53 seen

at these two timepoints in response to UV stimulation is less than that seen in N-MycER

cells induced with OHT. One possibility that could explain a lesser p53 response is that

the UV treatment conditions were insufficient to initiate a classical p53 mediated cell

cycle arrest response (review, (Harris and Levine 2005)). To address this possibility,

cells treated identically and in parallel were not harvested for protein at the given

timepoints, but were permitted to continue to grow. Morphological and cell number

changes were assessed 48 HRS after UV treatment. Cells UV treated for more than 10

seconds displayed a profound growth defect compared to untreated cells indicating that

the p53 response had indeed been elicited (data not shown).

RNAs were isolated from cells (described later) to be used for subsequent

expression analysis. To assess the quality of individual RNA preps, a small aliquot of

each was resolved under native conditions. All samples used in the subsequent

microarray analyses were monitored in this fashion, but some data are not shown. In

figure 2.4, overall RNA quality is interpreted as good because there is no evidence of

RNA degradation or genomic DNA contamination.

Before microarray experiments were initiated, we wanted to confirm that the

treatment conditions would cause the expected regulation of some previously described

Myc responsive genes (cad, cyclinD2 and hTert). Furthermore, we reasoned that we

could use these responses as a gauge to estimate the timing of the Myc response. We

performed reverse transcriptase-polymerase chain reaction (RT-PCR) analysis of a panel

20

V N- N+ ∆- ∆+ 0s 10s 20s 40s 0s 5s 20s 40s

Harvest 26h after UV

Harvest 10h after UV

α-FLAG

α-p53

Primary human fibroblasts

Figure 2.3 The p53 response to MycER activationA. Human primary foreskin fibroblast cells (BJ) expressing N-MycER, N-Myc(∆MBII)ER (∆), or control cells (V)) were treated with (+) or without (-) OHT. Whole cell lysates were made with RIPA buffer after 36hrs. B. Untransduced cells (BJ) were subjected to UV irradiation for indicated times. Whole cell lysates were extracted at indicated times. For samples in A and B, 70ug of total protein was resolved by PAGE and transferred to PVDF for western blot analysis of exogenous protein expression and p53 protein levels.

A B C A B C A B CVector NMycER N∆ER

0 Timepoints

- + - + - + - + - + - + - + - + - + - +NER ∆ER NER ∆ER Vec. NER ∆ER NER ∆ER NER

15’ 30’ 1hr 2hr 4hr

- + - +∆ER NER

- + - +∆ER NER

- + - +∆ER Vec

- + - +NER ∆ER

- + - +NER ∆ER

- + - +Vec. NER

- +∆ER

4hr 8hr 12hr 24hr 36hr 48hr

Figure 2.4 RNA quality assessmentHuman primary foreskin fibroblast cells (BJ) expressing N-MycER, N-Myc(∆MBII)ER (∆ER), or control cells (vector) were treated with (+) or without (-) OHT. Approximately 0.5ug of total RNA of each sample is resolved under native conditions (1% agarose gel in TAE). Due to low yield of RNA for some samples, less RNA was loaded for these samples.

21

of RNA samples harvested from the time course OHT induction experiment of the three

LCPX derivative cell lines. cDNAs were made using random primers during the RT step.

Using radio-labeled gene specific primers, expression level of specific genes was

determined by PCR of these cDNAs. Samples were resolved by native PAGE and

autoradiographs are presented in figure 2.5. To permit sample-to-sample comparison, we

used the GAPDH level in any given sample as a comparative control as has previously

been done (Bush et al. 1998) Consistently in all three Myc responsive genes, transcript

levels continued to rise in the time course until at least 36 HRS. Also, the response to N-

Myc(∆MBII)ER induction was always less than the response to the wildtype fusion

protein. Expression of hTERT in the early timepoints (15’ to 1HR) seem to have

unexplained responses; non-zero hTert levels are seen at 30’ in MycER cells not treated

with OHT and are also seen in N-Myc(∆MBII)ER cells at 1HR. Because of these

unexplained responses and the fact that both cad and cyclinD2 show no responsiveness at

the early timepoints, we chose not to use the early timepoints for subsequent analyses.

Based on these data, we chose to use 0 HR, 2 HRS, 4 HRS, 8 HRS, 12 HRS, 24 HRS, 36

HRS and 48 HRS timepoints after OHT treatment of N-MycER and N-Myc(∆MBII)ER

cells for microarray analysis. The following timepoints after OHT treatment of vector, or

(LCPX) cells were used: 0 HR, 8 HRS, 48 HRS. This type of validation analysis was

not performed on RNAs isolated from the cells infected with the constitutively expressed

constructs because these constructs were not new in the laboratory and have been

previously characterized. Some Myc and MycER responsive genes identified by

microarray analysis were confirmed by RT-PCR using the RNA preparations that were

22

- + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - +- + v wt ∆WT ∆ WT ∆ WT ∆ WT ∆ WT ∆ WT ∆ WT ∆ vec WT ∆ WT ∆ vec WT ∆

15’ 30’ 1hr 2hrs 4hrs 8hrs 12hrs 24hrs 36hrs 48hrs 0hr

CAD

CyclinD2

hTERT

GAPDH

OHT: - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - +- + v wt ∆WT ∆ WT ∆ WT ∆ WT ∆ WT ∆ WT ∆ WT ∆ vec WT ∆ WT ∆ vec WT ∆






CAD

CyclinD2

hTERT

GAPDH

OHT:

CAD

CyclinD2

hTERT

GAPDH

OHT:

Figure 2.5 RT-PCR analysis of cad, cyclinD2 and hTertHuman primary foreskin fibroblast cells (BJ) expressing N-MycER, N-Myc(∆MBII)ER, or control cells (LCPX) were treated with or without OHT for indicated times. Equal amounts of RNA prepared from these experiments were subject to RT-PCR for four transcripts (cad, cyclinD2, hTert and gapdh). One of the gene specific primers in the PCR reactions was end radiolabeled (32P). Reactions were resolved by native PAGE (TBE). Gels were dried and the autoradiograms are presented.

23

used for the microarray analysis from both time course and steady state experiments and

will be discussed later.

RNA harvested from these samples was first amplified using a modified Eberwine

protocol to synthesize a double stranded cDNA with a T7 promoter at the 3’end of each

mRNA (Eberwine 1996). Cy-5 labeled cRNA probes were synthesized by in vitro

transcription using T7 RNA polymerase. Total RNA was prepared from logarithmically

growing normal primary fibroblasts for a reference sample; references RNAs were

prepared separately for the time course experiments and the steady state experiments.

The reference sample was amplified in parallel with the experimental sample and Cy3-

labeled cRNA synthesized by in vitro transcription. The Cy-5 labeled and Cy-3 labeled

cRNAs were competitively hybridized to Agilent Technologies Human 1A version 2

oligo microarrays. Each microarray was scanned using an Axon Intruments GenePix

4000B microarray scanner and intensities collected by using GenePix5.0 software

(Molecular Devices Corp.). All data were stored in the UNC Microarray Database

(https://genome.unc.edu/).

To analyze the microarray data, we needed to decide how to normalize the spot

intensities. Two normalization protocols are commonly used (review, (Quackenbush

2002)). The total linear normalization assumes that the total intensity of all the probes in

one channel (i.e. one RNA sample) should equal the total intensity of all the probes in the

other channel (i.e. the competing RNA sample). If the sums of the intensities in the two

channels do not equal each other, then a normalization factor is typically applied globally

to one channel’s intensities such that the normalized total intensities from that channel

now equals the total intensity for the other channel. Locally weighted linear regression

24

(lowess) normalization can help reduce the effect of systematic intensity dependent

biases that have been reported to occur especially in low intensity spots. The lowess

normalization algorithm detects systematic deviations in the R-I (ratio-intensity) plot and

corrects these by applying a locally weighted linear regression (Quackenbush 2002).

While theoretical arguments can be made for choosing one normalization protocol over

another, the decision to use one protocol over another may be empirically guided using a

trusted assay for individual genes which are differently affected by different

normalization protocols. When the Myc time course data were normalized using the two

different normalization protocols, the data for some genes looked identical, but the

normalized data for some other genes from the two different protocols were remarkably

different. To address which protocol resulted in data that more accurately described the

underlying expression, we relied on RT-PCR to be the “referee” in some examples of

disagreements. Again, RT-PCR was designed to yield radiolabeled PCR products of

individual genes that could then be quantified using storage phosphor imaging

technology (Storm, Molecular Dynamics) (Figure 2.6-A) These data can then be

converted to a format that allows a simple visual comparison of RT-PCR data and

differently normalized microarray data (Figure 2.6-B,C). Four genes (map3k10, myl9,

ruvbl1, ebna1bp2) were chosen to be interrogated by RT-PCR as examples of the two

normalization protocols returning different expression data for these genes in the time

course experiment. For unexplained reasons, the steady state was not amenable to lowess

normalization. When the data were lowess normalized, obviously erroneous expression

data were returned. Therefore, for the steady state data, only the linearly normalized

microarray data is used for comparison in Figure 2.6-C.

25

A.

Vector N-MycER N-Myc(∆MBII)ERB.

1 2 31. LXSH2. N-Myc3. N-Myc(∆MBII)

C.

GAPDH

xx

EBNA1BP2

RuvBL1

MAP3K10

Myl9

Less cDNA used for this PCR. Do not use to normalize.

0h 24h 36h 0h 4h 8h 12h 24h 36h 48h V WT ∆MbII0h 4h 8h 12h 24h 36h 48h

Vector MycER MycER ∆MbII

GAPDH

xx

EBNA1BP2

RuvBL1

MAP3K10

Myl9



Vector MycER MycER ∆MbIIA.

Vector N-MycER N-Myc(∆MBII)ERB.

1 2 31. LXSH2. N-Myc3. N-Myc(∆MBII)

C.

GAPDH

xx

EBNA1BP2

RuvBL1

MAP3K10

Myl9




GAPDH

xx

EBNA1BP2

RuvBL1

MAP3K10

Myl9




GAPDH

xx

EBNA1BP2

RuvBL1

MAP3K10

Myl9




GAPDH

xx

EBNA1BP2

RuvBL1

MAP3K10

Myl9




Figure 2.6 Validation of normalized microarray data

26

Figure 2.6 Validation of normalized microarray dataA. RNA samples were subject to RT-PCR using a radioactively labeled gene specific PCR primer for each of the indicated genes. Reactions were resolved by native PAGE, gels were dried and individual bands were quantified using the Storm phosphor-imaging system (Molecular Dynamics). These data were normalized using the GAPDH level in each sample. The expression values were then normalized, or “zero-transformed” to the untreated control (in the timecourse experiments) or the vector control (in the steady state experiments). These values were then converted into log2 space and used to make the Java Treeview output in B and C. Linear or lowess normalized data that is presented was extracted from files in which the timecourse data had been normalized with the two different normalization protocols.

27

The linearly normalized microarray data for map3k10 in the time course

experiments suggests that expression is induced by N-Myc(∆MBII)ER, whereas if the

data were lowess normalized, the opposite conclusion is reached. By RT-PCR, the

map3k10 is repressed by N-Myc(∆MBII)ER suggesting that lowess normalization is

better suited for this probe. The other three genes were similarly investigated and we

found that the lowess normalization performed better at returning data most similar to the

RT-PCR data. Because we treated the RT-PCR as the benchmark for true transcript

levels, we chose to use lowess normalization for the time course experiments. The linear

normalization for these genes in the steady state experiments very accurately (when

compared to RT-PCR data) reports expression changes. As mentioned above, because

lowess normalization for the steady state state data leads to a normalization artifact, we

chose to use linear global normalization for the steady state data.

For the inducible Myc gene expression data, we performed a figure of merit

(FOM) analysis (TIGR MeV) on the filtered data (see Materials and Methods) to

determine the number of clusters that would provide a good fit for the expression data

using the k-means clustering algorithm (figure 2.7). For the steady state Myc gene

expression data, we used SAM (Significance Analysis of Microarrays, version 1.21) to

perform a two class, unpaired analysis with the wild type Myc arrays in one class and the

vector and mutant Myc arrays in another (Tusher et al. 2001). Genes showing

significantly differential expression identified by SAM were hierarchically clustered

using Cluster 3.0 (Eisen et al. 1998). Using Gene Ontology (GO) - TermFinder (version

0.61), a statistical tool to measure the overrepresentation of GO terms, clusters of

similarly responding genes were interrogated for overrepresented GO terms (Boyle et al.

28

Figure 2.7 Figure of Merit analysis of timecourse dataNormalized and zero transformed data were filtered for at least two data points across all experiments for a given probe to have a log2-ratio value greater than 0.8. FOM analysis is performed on these filtered data to determine an optimal number of clusters to choose for the subsequent k-means clustering analysis.

29

1 2 3 4

5 6 7 8

9 10 11 12

13 14 15

1 2 3 4

5 6 7 8

9 10 11 12

13 14 15

Figure 2.8 k-means clustering using 15 clusters of timecourse dataFiltered timecourse data were subject to k-means clustering using TIGR MeV. Euclidean distance was used as the clustering metric. A maximum of 50 iterations were permitted. Each expression graph represents the data for each probe that is in a cluster. Cluster number is given at the bottom right of each graph. The purple line in each graph represents the average response. Within each expression graph, the left graphs represent the vector timecourse data, the middle graphs represent the N-MycER timecourse data and the right graphs represent the N-Myc(∆MBII)ER data.

30

2004). The enrichment of GO terms among the clusters of Myc responsive genes

provides an unbiased assessment of the specific biological processes that are

nonrandomly represented in each cluster.

Overview of Myc induced Expression Profiles

Upon initial filtering (see Material and Methods), data from 1,631 probes (which

map to 1472 unique unigene clusters) were clustered using the k-means clustering

algorithm. Figure of Merit (FOM) analysis indicated that choosing 15 clusters for the

1,631 responding probes would provide a good fit of the expression data and that

increasing the number of clusters would provide limited improvement to the clustering

algorithm (Figure 2.7).

Line graphs of the 15 clusters generated by k-means clustering are presented in

Figure 2.8. Within each cluster, the left graphs represent vector time course data, the

middle graphs represent N-MycER time course data and the right graphs represent N-

Myc(∆MBII)ER time course data. This analysis attempts to put similarly responding

genes into common clusters. Due to the lack of time-dependent response of genes in

clusters 2 and 3, we suspect that these probe sets passed the initial filtering due to noisy

data points. Clusters 6 and 12 contain data that unexpectedly show a response to N-

MycER(∆MBII), but not the wildtype fusion protein. Of the 15 clusters, six clusters

(clusters 4, 5, 7, 9, 10 and 13) contained overrepresented biological process GO terms

(Figure 2.10). These clusters were also investigated for overrepresented component GO

terms. Some high level terms are omitted if more specific child terms are present. The

complete results are provided in supplementary material, Table 2.1 (sheet1). The filtered

time course data were also hierarchically clustered and are presented in Figure 2.9A.

31

MycER MycER∆MBIIEmptyVector

OHT (time):

1631 probes

A. B.

1608 probes

Emptyvector WT ∆MBII

1.43 2 - 31.4 2

Figure 2.9 Hierarchical clustering of timecourse and steady state dataA. Lowess normalized and zero transformed time course data were subject to a 1.7 fold change on at least two arrays across all data points filter. The data that passes this filter were hierachically clustered (average linkage). B. Linearly normalized and zero transformed steady state data were subject to two class unpaired analysis by SAM. The resulting probes differentiating the two classes (control and mutant versus wildtype Myc) are hierarchically clustered (average linkage).

32

Cluster 5

Cluster 9

Cluster 10

Cluster 13

Cluster 4

Cluster 7

Cluster 7

development 9.5E-07

morphogenesis 5.3E-06

cell communication 2.7E-05

organogenesis 7.8E-05

chemotaxis 1.1E-03

regulation of signal transduction 3.4E-03

signal transduction 7.2E-03

Cluster 4


skeletal development 7.7E-03

2.4E-04

nucleobase, nucleoside, nucleotide and nucleic acid metabolism

Cluster 13

5.6E-03biosynthesis

Cluster 10

3.5E-03protein biosynthesis

8.3E-04translation

5.1E-04macromolecule biosynthesis

Cluster 9

6.7E-03protein folding

1.2E-05RNA metabolism

4.7E-06RNA processing

Cluster 5

Figure 2.10 Overrepresented GO terms – time course experiments

33

Figure 2.10 Overrepresented GO terms – time course experimentsGenes in the clusters of the time course experiments were examined for enrichment of GO terms. Some GO terms from those clusters that were significantly enriched (Bonferroni corrected p-value < 0.01) in the “biological process” aspect are presented. See supplemental Table 2.1 for complete list of all enriched GO terms.

34

Emptyvector WT ∆MBII

Cluster of repressed genes


antigen processing, endogenous antigen via MHC class I 2.1E-04

reproductive physiological process 4.0E-04

organogenesis 6.6E-04

development 7.2E-04

6.3E-03cell cycle checkpoint

6.2E-03DNA replication initiation

6.1E-03tRNA metabolism

4.4E-05protein folding

3.6E-05DNA packaging

1.5E-07DNA-dependent DNA replication

3.3E-08rRNA processing

1.1E-08rRNA metabolism

4.4E-11ribosome biogenesis and assembly

1.9E-11cell proliferation

4.5E-13DNA replication

1.8E-14protein metabolism

6.4E-15DNA replication and chromosome cycle

3.1E-15mRNA processing

2.2E-15mRNA metabolism

4.4E-16nuclear mRNA splicing, via spliceosome

3.2E-16RNA splicing

1.1E-17cell cycle

2.2E-18DNA metabolism

2.0E-29protein biosynthesis

Cluster with upregulated gene expression

1608 probes

Fold change:

1.43 2 - 31.4 2

Figure 2.11 Overrepresented GO terms – steady state experiments

35

Figure 2.11 Overrepresented GO terms – steady state experimentsGenes in the upregulated and downregulated clusters of the steady state experiments were examined for enrichment of GO terms. Some GO terms that were significantly enriched (Bonferroni corrected p-value < 0.01) in the “biological process” aspect are presented. See supplemental Table 2.1 for complete list of all enriched GO terms.

36

SAM analysis identified 1608 differentially expressed probes in response to

steady state Myc expression with a median false discovery rate of 0.42%. These 1,608

probes map to 1,431 unique unigene clusters. The identified steady state Myc responsive

probes were hierarchically clustered using Cluster 3.0 (Figure 2.9B). Two distinct

clusters (a repressed and an induced) are evident. These two clusters were queried for

overrepresented GO terms and the representative GOTermFinder analysis results are

found in Figure 2.11. For complete results, see supplementary material, Table 2.1

(sheet2).

GO Term Overrepresentation

The GO terms that are overrepresented in some gene clusters (such as RNA

processing, protein biosynthesis and protein folding) strongly suggest that Myc

overexpression results in an increase in general protein synthesis. Genes involved in

ribosome biogenesis such as FBL (Fibrillarin), NPM1 (Nucleophosmin 1), NOL5A

(Nucleolar protein 5A) and EBNABP2 (EBNA1 binding protein 2) are upregulated by

Myc overexpression (Figure 2.14). Similarly, genes that encode products more directly

involved in protein biosynthesis such as 60S ribosomal protein L32, 39S ribosomal

protein L45, 40S ribosomal proteins S3 and S4 (Y isoform), 28S ribosomal protein S9

and MetAP2 (Methionine aminopeptidase 2) are also upregulated (Figure 2.14).

Several surprises in the data lay in the repressed genes. First, the number of

repressed genes forms a significant portion of the complete transcriptional response. In

the MycER experiments, approximately 48% of the response corresponds to repressed

genes. In the Myc steady state experiments, approximately 41% of the Myc response is

37

composed of repressed genes. Second, a substantial fraction of the repression response

initiated by MycER activation occurred as rapidly as the earliest induction response. For

example, PDGFR-α peptide (platelet derived growth factor receptor, alpha peptide 1),

DUSP1 (dual specific phosphatase 1) and ARHE (Ras homolog gene family member E)

show a repression response after only two hours of MycER activation (Figure 2.14).

Third, many of the repressed genes appear to be involved in communicating to other cells

or in responding to extracellular signals. Enriched GO terms in some of the repressed

clusters (in the case of the MycER experiments) and the whole repressed gene cluster (in

the steady state experiments) revealed a striking trend. Genes involved in

“development,” “organogenesis,” “morphogenesis” and “cell communication” are

repressed. A common element in these four GO terms is that each of these biological

processes involves the cell’s ability to recognize and communicate external cues to the

nucleus. Genes such as Rit1 (Ras-like GTP-binding protein), STAT1 (Signal transducer

and activator of transcription 1), EDG2 (LPA receptor), ITGB1 (Fibronectin receptor,

Integrin beta 1), SOS-1 (Son of sevenless protein homolog 1) are repressed and involved

in these processes (Figure 2.14). Finally, among the repressed genes, we noticed that

there was an overrepresentation of genes whose products are extracellular. In the steady

state experiment extracellular gene products are significantly overrepresented among the

repressed cluster (p-value=2.8x10-10). In the two repressed clusters 4 and 7 from the time

course experiments, extracellular gene products are similarly overrepresented in these

clusters (p-value=8x10-6 and p-value=5x10-11, respectively).

38

Reduced serum responsiveness

Provoked by these observations, we considered the possibility that Myc

overexpressing cells may have a compromised ability to respond to some external stimuli.

To investigate the responsiveness of Myc cells to some external cues, we assessed the

cells’ ability to respond to fetal bovine serum by measuring the phosphorylation of ERK

1/2. Control, MycER, or Myc(∆MBII)ER cells were either treated or not with OHT for

36 HRS. For the last 24 HRS of the OHT treatment period, cells were grown in serum

free medium. Cells were then treated with 10% serum and harvested 0’, 3’, 12’, 45’, or

120’ following serum stimulation. Cells expressing activated MycER showed reduced

serum responsiveness compared to control cells at the level of ERK 1/2 phosphorylation

of (Thr202/Tyr204) (Figure 2.12). Interestingly, the ERK 2 phosphorylation level in

MycER (+OHT) cells not stimulated with serum (0’) was higher than in control cells

[Vector (+OHT)] which were not stimulated with serum. Serum responsiveness was not

reduced in Myc(∆MBII) cells.

Comparison to other studies

Comparison of our study to previous studies that identified Myc responsive genes

underscores both the magnitude of the Myc response we have identified as well as the

differences inherent to the experimental approaches. Because we were interested in

exploring the total transcriptional response to Myc overexpression, we used two

experimental designs that reveal such responses over the course of 48 HRS (MycER) or

after a few weeks (steady state). To be as inclusive as possible for comparisons to other

studies, we considered the Myc Target Gene database as a repository of previously

described Myc responsive genes (Zeller et al. 2003). The Myc Target Gene database is

39

Vec

tor

Myc

ERM

ycER

(∆M

BII)

Vec

tor

Myc

ERM

ycER

(∆M

BII)

-OH

T+O

HT

Phos

pho-

ERK

:

Tota

l ER

K:

vectorMycERMycER(∆MBII)

FLA

G

0’0’

0’0’

0’0’

Seru

m:

Vec

tor

Myc

ERM

ycER

(∆M

BII)

Vec

tor

Myc

ERM

ycER

(∆M

BII)

-OH

T+O

HT

Phos

pho-

ERK

:

Tota

l ER

K:

vectorMycERMycER(∆MBII)

FLA

G

0’0’

0’0’

0’0’

0’0’

0’0’

0’0’

Seru

m:

A.

B.

Figu

re 2

.12

Ser

um r

espo

nsiv

enes

sH

uman

prim

ary

fore

skin

fibr

obla

st c

ells

(BJ)

exp

ress

ing

N-M

ycER

, N-M

yc(∆

MB

II)E

R, o

r con

trol c

ells

(LC

PX) w

ere

treat

ed

with

or w

ithou

t OH

T fo

r 36h

rs a

nd w

ere

seru

m st

arve

d fo

r 24h

rs.

Then

, cel

ls w

ere

treat

ed st

imul

ated

with

10%

FB

S in

D

MEM

. W

hole

cel

l ext

ract

s wer

e pr

epar

ed 0

’, 3’

, 10’

, 30’

or 9

0’af

ter s

erum

stim

ulat

ion

with

RIP

A b

uffe

r. L

ysat

esw

ere

quan

titat

ed.

Equa

l am

ount

s of l

ysat

efr

om e

ach

sam

ple

wer

e lo

aded

in tw

o ge

ls in

par

ralle

land

reso

lved

by

PAG

E. U

pon

trans

fer t

o PV

DF,

par

alle

l blo

ts w

ere

prob

ed w

ith p

hosp

hory

latio

nsp

ecifi

c ER

K1/

2 an

tibod

ies o

r tot

al E

RK

1/2

ant

ibod

ies.

B.

Lysa

tesp

repa

red

from

thes

e ce

lls im

med

iate

ly b

efor

e O

HT

treat

men

t wer

epr

obed

for e

xpre

ssio

n of

the

trans

gene

susi

ng

FLA

G a

ntib

odie

s.

40

Myc target gene database1389 genes

MycER responsive 1472 genes

Myc (steady state) responsive1431 genes

98

406

267

256

Figure 2.13 Venn diagram: Myc responsive genes in different studiesMyc responsive genes in our studies and those found in the Myc cancer database were mapped to unique unigene cluster ID’s. The total number of target genes from the Myc database is the number of genes in the database that map to unique unigene cluster ID’s and that had a probe on the agilent arrays.

41

populated by Myc responsive genes that have been identified by SAGE analysis, DNA

microarray analysis, subtractive hybridization analysis, ChIP analysis, northern/western

blot analysis or RT-PCR analysis. To compare MycER and the steady state Myc studies

with a current compilation of Myc targets in the Myc Target Gene database, we mapped

Myc responsive genes in our studies and those found in the Myc cancer database to

unigene cluster ID’s (locus link ID’s of database genes, provided by C. Dang,

supplemental table 2.2). The total number of target genes from the Myc database is the

number of genes in the database that map to unique unigene cluster ID’s and that had a

probe on the agilent arrays. Similarly, the number of MycER responsive or Myc

responsive genes is the number of responsive probes that map to unique unigene cluster

ID’s. In figure 2.13, we express the comparison of the data in Venn diagram form.

When individually compared to the Myc database, the MycER responsive genes overlap

with 256 genes out of 1389 database genes and the steady state Myc responsive genes

overlap with 267 out of 1389 database genes. Surprisingly, only 98 genes are common to

all three groups. Thus, our analysis reveals a large set of previously unrecognized Myc

responsive genes.

Definition of the Myc gene expression signature

We have examined the transcriptional response to Myc overexpression by using two

different and independent Myc overexpression experimental designs. As described above,

we first used an inducible Myc construct to monitor cellular transcriptional changes over

the course of 48 HRS. We then monitored transcriptional changes after three weeks of

overexpression of a constitutively overexpressed MYC construct. Those genes

responsive to Myc overexpression in both experimental designs are classified as the core

42

Myc

ERM

ycER

∆MB

II

OH

T (ti

me)

:

Vec

tor

Vec

tor W

T∆MB

II

Fibr

alla

rin

(FB

L)

Nuc

leol

arPr

otei

n 5A

(NO

L5A

)

EB

NA

1 B

indi

ng P

rote

in 2

(EB

NA

1BP2

)

Ras

-like

GT

P bi

ndin

g pr

otei

n (R

IT1)

Plat

elet

der

ived

gro

wth

fact

or r

ecep

tor

(PD

GFR

-α)

Nuc

leop

hosp

min

(NPM

)

Rib

osom

al p

rote

in S

3

Dua

l spe

cific

pho

spha

tase

1 (D

USP

1)

Fibr

onec

tinre

cept

or (I

TG

B1)

Orn

ithin

ede

carb

oxyl

ase

(OD

C)

Met

hion

ine

Am

inop

eptid

ase

2(M

etA

P2)

Rib

osom

al p

rote

in, S

9

Mito

chon

dria

l rib

osom

al p

rote

in, L

32

Rib

osom

al p

rote

in S

4

Mito

chon

dria

l rib

osom

al p

rote

in, L

45

Ras

hom

olog

gen

e fa

mily

mem

ber

E

(AR

HE

)L

PA r

ecep

tor

(ED

G2)

Figu

re 2

.14

The

Myc

gen

e ex

pres

sion

sign

atur

e

43

Figure 2.14 The Myc gene expression signatureExpression data for probes that are identified as responsive to both MycER activation and steady state Myc overexpression are hierachically clustered. The expression response of these probes is defined as the Myc signature. The signature probes total 428 agilent probes which can be mapped to 394 unique unigene clusters.

44

transcriptional response genes, or the Myc gene expression signature. As microarray

technology permits the measurement of global transcriptional changes, the apparent

response to Myc overexpression may be confounded by many variables such as construct

type, extent of overexpression, time of expression, natural variation of cells in culture,

cell density etc. Since we took two different experimental approaches, we reasoned that

the genes common to the Myc response in both experimental approaches would serve as a

conservative estimate of a core Myc gene expression signature.

The core Myc gene expression signature, or simply the Myc signature, consists of

428 agilent probes which map to 394 unique unigene clusters (Figure 2.14). Ninety one

percent (91%) of these genes have the same direction of response in both studies.

Surprisingly, 9% of the signature contains genes which are induced in one study and

repressed in the other. We measured GO term overrepresentation among the clusters of

genes which display a common direction of response and found that the overall nature of

the induced genes and repressed genes is very similar to that found in the individual

studies. Hence, we concluded that our method of determining the Myc signature did not

result in a grossly skewed subset of Myc responsive genes.

Analysis of potential Myc binding sites in Myc signature genes

The promoters of bona fide Myc target genes tend to have two common

characteristics (Fernandez et al. 2003). First, Myc binding sites are found near the

transcriptional start site; these binding sites are commonly located 3’ of the

transcriptional start site. Second, the promoter regions are often rich in CpG

dinucleotides. Therefore, we wanted to evaluate whether the Myc binding site is

overrepresented in the region proximal to the transcriptional start site of Myc signature

45

Genes Promoter proximal E-box p-value

Whole array 16.6% N/A

Repressed in Myc signature 14.3% not significant

Induced in Myc signature 32.3% 1.4x10-8

Complete Myc signature 24.8% 3.2x10-5

MycER MycER∆MBIIVectorVector

WT∆MBII

repressed

induced


WT∆MBII

repressed

induced


WT∆MBII

repressed

induced


WT∆MBII

repressed

induced

The Myc signature

Figure 2.15 E-box overrepresentationThe percentage of genes within an indicated group of genes that contained at least one E-box within 500bp of the transcriptional start site was determined. The statistical significance of the overrepresentation of an E-box in a group of genes when compared to the background representation of E-boxes in genes of the whole array (Agilenthuman 1A) is calculated by the hypergeometric distribution and presented as a p-value. (N/A, not applicable).

46

genes. To do this, we first retrieved sequence (+/- 500bp) flanking the transcriptional

start site (TSS) of all the genes on the microarray. This step was performed using

PromoSer (Halees et al. 2003). These sequences were then evaluated for the presence of

at least one E-box (CACGTG). The percentage of total genes on the array that contained

an E-box within 500bp of the TSS was determined to be 16.6% (Figure 2.15). When we

perform a similar analysis on the total Myc signature gene list, 24.8% of the genes

contain E-boxes. Assuming a hypergeometric distribution, this difference is statistically

significant (p-value= 3.2x10-5). When we examine the upregulated genes, the percentage

of genes with promoters containing E-boxes rises to 32.3% (p-value=1.4x10-8). The

promoters of the Myc signature repressed genes, however, had E-boxes in only 14.3% of

the genes which is not significantly different than the background representation level.

This supports the idea that Myc positively regulates gene expression of direct targets by

binding Myc binding sites. However, the repressive effects of Myc on direct targets are

likely to occur via a different mechanism.

Systematic analysis of the Myc Signature in a breast tumor profiling dataset

The role of Myc in breast cancer has been long studied. Reports that come to

different conclusions exist in the literature (review, (Liao and Dickson 2000)).

Depending on the study, Myc amplifications are found in anywhere from 1% to 94% of

all breast cancers. Equally confusing, both good and poor prognoses have been

correlated with high Myc expression. In this light, we wanted to determine if the core

Myc signature is found in any subgroup of breast tumors and if that group would exhibit

any correlations with increased or decreased survival. RNA from biopsy material from

104 breast cancer patients was harvested and analyzed on Agilent’s human 1A version

47

microarrays. Expression profiling of the breast tissues was performed in the Perou lab

at UNC and is presented in a manuscript which is under review (Hu et al. 2005).

Using only the tumor data that corresponds to the Myc signature gene list, we

hierarchically clustered the tumor samples in two dimensions and asked whether the

clustering of the tumors driven by the Myc signature would resolve the different subtypes

of breast tumors. The Myc signature does not strictly resolve all four breast tumor

subtypes (Basal, Luminal, HER2+, and Normal-like) into four clusters as the intrinsic

gene list does (Figure 2.16). However, the Myc signature does cluster every basal tumor

into one cluster (the left (purple) branch of the array dendogram) which is inhabited by

only three other previously classified non-basal tumor samples (two are HER2+ and one

is a luminal tumor) (Figure 2.17).

The observation that the Myc signature gene list can drive the clustering of the

basal tumors into one cluster gives a prediction that the basal tumors have either the

highest or the lowest Myc levels among the tumors and that the gene expression profile

of the Myc signature is very similar to or is almost the inverse of the Myc signature. To

explore relative differences in c-Myc RNA levels across the tumors, we displayed the

median centered data for the Myc probe in histogram form below the tumor samples

(Figure 2.16). Among the tumor samples, the basal tumors tend to have the highest Myc

levels. Furthermore, the general expression pattern of the basal tumors of the Myc

signature genes is strikingly similar to the fibroblast response to Myc overexpression.

Interestingly, some of the normal breast tissue samples which are in the right (blue)

branch of the array dendrogram have relatively high Myc expression and display

concordance with only part of the Myc signature (like some of the ribosomal genes), but

48

Myc

sign

atur

eB

reas

t tis

sue

expr

essi

on p

rofil

es

Med

ian

cent

ered

c-M

yc R

NA

leve

ls:

Figu

re 2

.16

The

Myc

gen

e ex

pres

sion

sign

atur

e in

bre

ast t

umor

tiss

ue

49

Figure 2.16 The Myc gene expression signature in breast tumor tissue The Myc signature gene list was extracted from a breast cancer profiling dataset. These data for these probes were hierarchically clustered in two dimensions (genes and arrays). The order of breast tissue arrays (right) was preserved before merging with Myc overexpressing fibroblast data (left). The merged data set was hierarchically clustered in one dimension (genes only) to give the presented format. The median centered c-Myc RNA levels were extracted from the breast tissue expression profiling study.

50

A.1 2 3 Sum

Lum (43) 1 22 20 43

Normal (6)

0 6 6

Basals(28)

28 28

Her2 (16) 2 7 7 16

Sum 31 29 33

93

unclass 2 7 2 11

104

B.

0

0.2

0.4

0.6

0.8

1

1.2

0 50 100 150

RFS months

Prob

abili

t Censored

1

2

3

Figure 2.17 Cluster and survival analysis

51

Figure 2.17 Cluster and survival analysisColored dendrogram branches from Figure 2.15 correspond to the colors used in thisfigure. A. Analysis of the type of breast tissues in each left (purple), middle (yellow), and right (blue) array dendrogram branches. B. Kaplan-Meier survival plots of groups of patients whose breast tissues were profiled. The groups of patients for this analysis were defined by the clustering analysis in figure 2.16.

52

they lack concordance with the rest of the Myc signature expression cluster that is seen

in the basal tumors. Indeed, they cluster together, but quite distantly from the basal

cluster. Interpretation of the high Myc levels in the normal samples is confounded by the

fact that normal breast tissue samples are generally more heterogeneous than tumor

samples; adipose tissue forms a greater percentage of normal tissue than of resected

tumor tissue.

Quantifiable presence of the Myc signature

To quantify the resemblance between the Myc signature response and the

expression profile of an individual tissue, we first calculated a Myc signature centroid to

which the resemblance of each expression profile is calculated as a Pearson’s correlation.

A similar strategy was previously employed to calculate the presence of the “wound

signature” in tumor expression profiles (Chang et al. 2005). In our study, we use this

correlation value as a means to describe the presence of the Myc signature in a given

sample. However it is difficult to decide a threshold correlation value above which

would classify a sample as definitively displaying the Myc signature. In comparison, the

absolute value of a correlation score describing the presence of the wound signature in a

tumor sample of greater than 0.15 had substantial predictive clinical consequences

(Chang et al. 2005). In figure 2.18, the Pearson’s correlation between the Myc centroid

and breast tumor expression profiles is presented. The left, middle and right clusters

display average correlation scores of 0.17, 0.04, and -0.25, respectively. In the left

cluster, nearly all the tumors had a positive correlation score and some tumors had

correlation scores greater than 0.5. This analysis allows us to quantify the resemblance of

the expression profile of each cluster to the Myc signature.

53

Myc

sign

atur

eB

reas

t tis

sue

expr

essi

on p

rofil

es

Pear

son

corr

elat

ion

to M

yc si

gnat

ure

cent

roid

:

Figu

re 2

.18

Qua

ntifi

catio

n of

the

Myc

sign

atur

e in

bre

ast t

umor

tissu

e

54

Figure 2.18 Quantification of the Myc signature in breast tumor tissueThis figure contains the same data as in Figure 2.14, however, pearson correlation values are presented describing the resemblance of an individual expression profile to the Myc signature centroid.

55

A.

B.

Figure 2.19 c-Myc protein in normal and basal breast carcinoma tissueAntibody staining of c-Myc in paraffin sections of normal breast tissue (A) and basal breast carcinoma tissue (B).

56

Immunohistochemical analysis of Myc in breast carcinoma tissue

While data centering can draw our attention to differences in RNA levels, the

centered data will not describe absolute levels or even which cells in a heterogeneous

sample actually are responsible for the high (or low) expression levels of a specific gene.

Therefore, we performed immunohistochemistry (IHC) on 13 basal tumor tissue sections

using antibodies against Myc. Twelve of the 13 basal tumors were positive for Myc.

Representative IHC images of normal breast tissue with low Myc expression and basal

tumor tissue with high Myc levels are shown in figure 2.19. While 92% of the basal

tissues tested were positive, only 41% (9 out of 22) of the non-basal tumor tissues tested

displayed high Myc levels by IHC. This further supports the conclusion that the Myc

gene expression signature is a common, and perhaps critical, characteristic of basal breast

cancer tissue.

Analysis of the Myc Signature in other human tissue expression profiling datasets

We wanted to know whether the Myc signature list can be used to identify a Myc

response in other datasets. To this end, we performed a similar analysis of several

publicly available expression profiling datasets. Because other datasets use different

arrays than in our study, we first mapped the Myc signature and the others’ datasets to

unigene cluster IDs. Mean or median centered tumor data corresponding to the Myc

signature were hierarchically clustered in two dimensions. Our study was then linked to

the clustered tumor data and reclustered in one dimension (genes). We examined the

Myc signature in other datasets [Acute lymphoblastic leukemia (Cario et al. 2005), head

and neck squamous cell carcinoma (HNSCC) (Chung et al. 2004), various normal human

57

Met

asta

tictis

sue

from

pro

stat

e ca

ncer

pat

ient

Pros

tate

tum

or ti

ssue

Nor

mal

pro

stat

e tis

sue

Mea

n ce

nter

ed c

-Myc

leve

ls:

Lapo

inte

et a

l 200

5Fi

gure

2.2

0 T

he M

yc g

ene

expr

essi

on si

gnat

ure

in p

rost

ate

tissu

es

Myc

sign

atur

e

58

Figure 2.20 The Myc gene expression signature in prostate tissues The Myc signature gene list was converted to unigene cluster IDs. This list was extracted from a prostate tissue profiling dataset (Lapointe et al. 2004). The data for these probes were hierarchically clustered in two dimensions (genes and arrays). The order of prostate tissue arrays (right) was preserved before merging with Myc overexpressing fibroblast data (left). The merged data set was hierarchically clustered in one dimension (genes only) to give the presented format. The mean centered c-Myc RNA levels were extracted from the prostate tissue expression profiling study.

59

tissues (Shyamsundar et al. 2005), breast carcinoma (van 't Veer et al. 2002), lymphoma

tissues (Lossos et al. 2002) and normal and malignant prostate tissues

(Lapointe et al. 2004)]. The Myc signature profile is expressed to varying degrees in

these datasets. In the ALL and normal human datasets, there was little co-occurrence

between high relative Myc levels and the Myc Signature profile, whereas the HNSCC,

breast carcinoma, FL/DLBCL lymphoma, and prostate tissue datasets showed greater

concordance (data not shown).

Prostate tumors that display the Myc signature (indicated as the red branch of the

tumor dendogram in figure 2.20) cluster together and generally have the highest c-Myc

RNA levels (Lapointe et al. 2004). Normal prostate tissues also cluster together, but do

not display the Myc signature. All but one of the normal tissues and all but one of the

metastasis tissues clustered separately into two distinct clusters. The cluster of tumors

that is composed exclusively of metastasis tissue displayed only part of the Myc signature.

Intriguingly, this metastasis tissue cluster did not display consistently high relative Myc

levels.

A recent study profiling normal tissues from various sites in the human body

reported that clustering of expression data leads to effective proximal clustering of

common tissues (Shyamsundar et al. 2005). We reasoned that proliferative tissues would

have the highest MYC RNA levels and would display the Myc signature. Upon two-way

hierarchical clustering of the Shyamsundar data corresponding to the Myc signature

genes, we found that some tissues such as tonsil, lymph node, gall bladder display the

Myc signature (Pearson’s correlation to the Myc signature > 0.2) (Figure 2.21). Other

tissues that are contain some highly proliferative cells such as lung, stomach, colon and

60

Mea

n ce

nter

ed c

-Myc

RN

A le

vels

:

Myc

sign

atur

eN

orm

al ti

ssue

exp

ress

ion

prof

iles Sh

yam

sund

aret

.al2

005

Figu

re 2

.21

The

Myc

gen

e ex

pres

sion

sign

atur

e in

nor

mal

hum

an ti

ssue

s

61

Figure 2.21 The Myc gene expression signature in normal human tissuesThe Myc signature gene list was converted to unigene cluster IDs. This list was extracted from a normal human tissue profiling dataset (Shyamsundar et al. 2005). The data for these probes were hierarchically clustered in two dimensions (genes and arrays). The order of normal tissue arrays (right) was preserved before merging with Myc overexpressing fibroblast data (left). The merged data set was hierarchically clustered in one dimension (genes only) to give the presented format. The mean centered c-Myc RNA levels were extracted from the prostate tissue expression profiling study.

62

Myc

cen

troid

cor

rela

tion

c-M

yc R

NA

Figu

re 2

.22

The

Myc

sign

atur

e vs

. c-M

yc le

vels

in n

orm

al h

uman

tiss

ues

To c

ompa

re th

e M

yc si

gnat

ure

with

c-M

yc le

vels

, pea

rson

corr

elat

ions

bet

wee

n th

e M

yc si

gnat

ure

cent

roid

and

each

no

rmal

hum

an ti

ssue

exp

ress

ion

prof

iles w

as c

alcu

late

d. T

hese

cor

rela

tion

valu

es w

ere

plot

ted

from

hig

hest

to lo

wes

t al

ong

with

the

corr

espo

ndin

g ce

nter

ed c

-Myc

val

ues.

63

esophagus did not display the Myc signature. Complete listing of array labels from

figure 2.21 are found in supplemental table 2.3. Particularly surprising was that c-Myc

levels from these tissues did not mimic the Pearson’s correlation values between the Myc

signature and each tissue’s expression profile. An illustration of this phenomenon is

found in Figure 2.22 in which the Pearson’s correlation value of each profile has been

sorted from highest to lowest and graphed along with the accompanying c-Myc levels.

As discussed later, one explanation for this result is that the signature response to

deregulated Myc expression is substantially different from the signature response to

normal Myc expression.

To explore the Myc signature in the development of a malignancy known to

involve Myc deregulation, we used a dataset that profiled tissue from patients of different

stages of disease progression. Follicular lymphoma (FL) is an indolent malignancy that

has been strongly associated with deregulated expression of BCL2 family genes, but

deregulation of MYC expression is uncommon. FL can transform to a very aggressive

malignancy called diffuse large B-cell lymphoma (DLBCL) (Acker et al. 1983; Horning

and Rosenberg 1984). Mutations that deregulate MYC expression are among the most

common that are believed to mediate the transformation from FL to DLBCL (Yano et al.

1992). Biopsy specimens from twelve patients who were diagnosed with follicular

lymphoma (FL) were expression profiled (Lossos et al. 2002). These same patients were

treated, but later returned to the clinic with the more aggressive DLBCL diagnoses.

Biopsy specimens from these patients after the FL to DLBCL transformation were also

profiled. In addition to these 24 samples in the Lossos dataset, the study included 11

samples from other patients whose primary diagnosis was DLBCL rather than FL, 3

64

Initi

ally

dia

gnos

ed w

ith F

ollic

ular

lym

phom

a (F

L)FL

pat

ient

s with

dis

ease

pro

gres

sion

to d

iffus

e la

rge

B-c

ell l

ymph

oma

(DLB

CL)

D

LBC

L at

initi

al d

iagn

osis

DLB

CL

cell

lines

Ger

min

al B

-cel

l lym

phoc

ytes

Loss

oset

.alP

NA

S (2

002)

Initi

ally

dia

gnos

ed w

ith F

ollic

ular

lym

phom

a (F

L)FL

pat

ient

s with

dis

ease

pro

gres

sion

to d

iffus

e la

rge

B-c

ell l

ymph

oma

(DLB

CL)

D

LBC

L at

initi

al d

iagn

osis

DLB

CL

cell

lines

Ger

min

al B

-cel

l lym

phoc

ytes

Loss

oset

.alP

NA

S (2

002)

Figu

re 2

.23

The

Myc

gen

e ex

pres

sion

sign

atur

e in

B-c

ell l

ymph

omas

Myc

sign

atur

e

65

Figure 2.23 The Myc gene expression signature in B-cell lymphomasThe Myc signature gene list was converted to unigene cluster IDs. This list was extracted from lymphoma profiling dataset (Lossos et al. 2002). The data for these probes were hierarchically clustered in two dimensions (genes and arrays). The order of lymphoma arrays (right) was preserved before merging with Myc overexpressingfibroblast data (left). The merged data set was hierarchically clustered in one dimension (genes only) to give the presented format. The mean centered c-Myc RNA levels were extracted from the prostate tissue expression profiling study.

66

samples of DLBCL cell lines, and 1 sample of isolated germinal center B-lymphocytes.

This dataset is particularly useful because it contains data from matched patients before

and after a suspected MYC deregulatory event. It is important to recognize the “before”

samples were not disease-free samples, but rather were tissues from patients with FL.

In Figure 2.23, the Myc signature genes are examined as above. All the FL

samples cluster together, but away from the closely clustered DLBCL samples. As

expected, DLBCL samples tend to have higher c-Myc levels than the FL samples. When

we examine the relative expression level of the Myc signature genes, we found that many

of the genes which are induced by Myc in fibroblasts show high relative expression in the

DLCBL samples. Few Myc signature genes (which are repressed in fibroblasts) are

relatively repressed in DLBCL. Similarly, another group of Myc signature genes (which

are induced in fibroblasts) show discordant regulation in the FL to DLBCL

transformation. Speculation on the inconsistencies is discussed later.

Analysis of the Myc Signature in a serum response dataset

In vivo, cells experience serum stimulation normally only in the context of local

injury. The gene expression program in response to serum exhibits many features of the

wound healing response (Chang et al. 2004). Motivated by the observation that wound

healing and tumor growth share many common physiological characteristics (review,

(Bissell and Radisky 2001)), Chang and colleagues showed that the expression of the

serum response in cancer cells is a powerful predicator of cancer patient survival and

metastasis. In tissue culture, Myc transformed cells exhibit reduced dependence on

serum for rapid proliferation and survival (Keath et al. 1984). It is commonly believed

that Myc enables cells to develop such independence by directly or indirectly inducing or

67

repressing those genes that are normally regulated by serum stimulation. To examine

how the Myc signature genes are affected by serum stimulation, we utilized a publicly

available serum stimulation expression profiling dataset of normal human fibroblasts

(Chang et al. 2004). When our study is compared to the serum stimulation study, we find

that some of the Myc responsive genes are, expectedly, also serum responsive (Figure

2.24A) There are, however, cohorts of Myc responsive genes which do not respond to

serum (i.e. REPIN1, LARS, SARS2) or, surprisingly, respond oppositely (i.e. TFPI2,

ITGA2, VEGFC, F3) (Figure 2.25) to serum stimulation. Many of the products of

concordant genes upregulated by Myc and serum are involved in formation and

maturation of ribosomes (NCL, FBL, EBNA1BP2, NOL5A). We then examined the

regulation of the core serum response genes (as defined by Chang and colleagues) in the

Myc response. Again, there is marked discordance in some of the genes regulated by

serum and Myc (Figure 2.24B). These observations may serve to dispel the notion that

Myc short circuits the need for serum by simply regulating all serum responsive genes in

the same direction.

68

Myc

ERM

ycER

∆M

BII

time:V

ecto

rSe

rum

resp

onse

Cor

e Se

rum

Res

pons

ive

gene

s in

the

Myc

resp

onse

Myc

ERM

ycER

∆M

BII

time:V

ecto

rV

ec WT∆

Seru

m re

spon

seV

ec WT∆

Myc

sign

atur

e ge

nes i

n th

e se

rum

resp

onse

378

gene

s

307

gene

s

Cha

ng e

t al 2

004

A.

B.

Myc

ERM

ycER

∆M

BII

time:V

ecto

rSe

rum

resp

onse

Cor

e Se

rum

Res

pons

ive

gene

s in

the

Myc

resp

onse

Myc

ERM

ycER

∆M

BII

time:V

ecto

rV

ec WT∆

Seru

m re

spon

seV

ec WT∆

Myc

sign

atur

e ge

nes i

n th

e se

rum

resp

onse

378

gene

s

307

gene

s

Cha

ng e

t al 2

004

A.

B.

Figu

re 2

.24

Ser

um st

imul

atio

n vs

. Myc

ove

rexp

ress

ion

69

Figure 2.24 Serum stimulation vs. Myc overexpressionA. The Myc signature gene list was converted to unigene cluster IDs. The data for this list were extracted from serum stimulation dataset (Chang et al. 2004) and merged with the Myc signature responses. The data for these probes were hierarchically clustered in one dimension (genes). Upon serum stimulation, samples were harvest from 0hrs to 36hrs . B. The inverse procedure was performed. The data for those genes defined as core serum responsive (CSR) were extracted to from the both studies and merged.

70

RE

PIN

1L

AR

SSA

RS2

NC

L

FBL

EB

NA

1BP2

TFP

I2IT

GA

2V

EG

FC

F3

Myc

ERM

ycER

∆MB

II

OH

T/se

rum

(tim

e):V

ecto

rEm

pty

vect

or WT∆M

BII

Seru

m re

spon

se

NO

L5A

Cha

ng e

t al.

2004

RE

PIN

1L

AR

SSA

RS2

NC

L

FBL

EB

NA

1BP2

TFP

I2IT

GA

2V

EG

FC

F3

Myc

ERM

ycER

∆MB

II

OH

T/se

rum

(tim

e):V

ecto

rEm

pty

vect

or WT∆M

BII

Seru

m re

spon

se

NO

L5A

Cha

ng e

t al.

2004

Figu

re 2

.25

Myc

sign

atur

e in

the

seru

m r

espo

nse

This

is th

e sa

me

data

as i

n Fi

gure

2.2

1A w

ith so

me

gene

labe

ls.

71

Discussion

The fact that nearly in all mammalian cells, proliferation is contingent on the

expression of at least one Myc family protein is suggestive of the central role Myc has in

cell proliferation. Given its central role in proliferation, it is not surprising that aberrant

or deregulated Myc expression can lead to a malignant transformation of cellular growth.

Numerous high-throughput studies querying Myc regulated genes have recently

been published (Coller et al. 2000; Guo et al. 2000; Schuhmacher et al. 2001; Menssen

and Hermeking 2002; Watson et al. 2002; O'Connell et al. 2003; Mao et al. 2004;

Marinkovic et al. 2004). Perhaps surprisingly, little overlap in target genes is found

among all the studies. A trivial but important reason for part of the non-overlap is that

the genes on the microarrays were different from study to study. As newer microarray

technologies permit complete or near complete genome wide inquiry, this problem may

be resolved. Numerous other possible explanations stem from methodological

differences including choice of cell type, the levels of Myc overexpression, time of

overexpression, and the choice of MYC family member as the exogenous transgene.

While many studies have focused on the discovery of Myc target genes, few have

examined the complete cellular transcriptional changes induced by Myc. A recent study

of gene expression in a murine model of prostatic intraepithelial neoplasia (mPIN)

created by targeting MYC overexpression to the prostate has defined a prostate cancer

specific Myc induced gene expression signature (Ellwood-Yen et al. 2003). In another

study, analysis of gene expression profiles of neuroblastomas identified a cooperation

between high N-Myc levels and high levels of another oncogene, H-Twist, that functions

to combat N-Myc’s proapoptotic effects (Valsesia-Wittmann et al. 2004). In these

72

studies, comparison between tissues with low Myc levels to tumor tissues with high

Myc levels was used to define the Myc dependent gene expression profile. Because the

comparisons are to tumors with high Myc expression - which are likely to have suffered

multiple other genetic lesions - these strategies reveal the transcriptional profile after all

genetic lesions have occurred. In other words, these studies identify changes in gene

expression induced by Myc overexpression plus those resulting from altered activity of

additional oncogenes or tumor suppressor genes. Recently, a Myc profiling study was

performed in rat cells in which both c-Myc alleles were disrupted (O'Connell et al. 2003).

In this study, exogenous MYC transgenes were expressed to examine the effect of

reintroducing Myc into a Myc null background. These experiments, however, were

performed in an immortalized cell background in which there are likely to be unknown

genetic abnormalities. A more fundamental Myc gene expression profile is likely to be

broadly relevant to different types of Myc-driven tumors and should facilitate elucidation

of the complex cellular responses that mediate the biological activities of the Myc protein.

We employed two primary cell culture models of Myc overexpression to define a

fundamental transcriptional gene expression profile that probably results from a single

molecular perturbation. The gene expression profiles of clinical samples are

compilations of the gene expression changes due to many environmental and genetic

perturbations. It is therefore difficult to ascribe, a priori, the effect any individual

perturbation has on gene expression. If, however, a gene expression program initiated by

a specific perturbation is known, we can use this knowledge to search for an individual

program in clinical expression profiles certain to reflect the transcriptional program of

many perturbations.

73

The expression profiles initiated by Myc overexpression in fibroblast cells

revealed that many genes encoding products which are involved in a variety of processes

are affected. When we examined the expression of these Myc responsive genes in human

tumor expression profiling datasets, we found that some tumors display the Myc

signature response. In the case of a breast cancer dataset, almost all the tumors that

displayed the Myc signature were from the basal-like subtype of breast cancer and the

patients with these tumors had the worst prognosis. This suggests a common underlying

biological program in effect in both the basal tumors and the Myc induced fibroblast cells.

From these data, it is not formally possible to conclude that higher Myc levels or

expression of the Myc signature is essential for the pathology witnessed in the patients

with basal breast tumors. Given the oncogenic properties of Myc, however, it is likely

that this expression profile is not inconsequential, but defines critical molecular events

underlying the pathological phenotype.

When the Myc signature was examined in other publicly available human tissue

expression profiling datasets, we find that some cancer datasets have prominent clusters

of tumors which express the Myc signature. In the prostate tissue dataset, we were able

to identify a sub-group of primary tumors that displayed the Myc signature. All the

metastatic tissue samples clustered together and displayed part of the Myc signature. In

other cancer datasets, we were unable to identify a true Myc signature. It is likely that

these datasets simply do not have a subset of tissues/tumors that express a coherent Myc

signature. While we have noticed that the Myc signature as defined in fibroblast is useful

in identifying the Myc signature in clinical samples of different cell types, it is possible

that the transcriptional response to MYC overexpression may widely vary from one cell

74

type to another. This could explain why in some datasets, the fibroblast Myc signature

is not found. Another unlikely possibility is that every tissue/tumor sample expressed the

Myc signature. Our methodology would not be able to recognize that scenario because

mean or median centering microarray data allows us to highlight differences within an

expression profiling dataset.

The profiling study of the FL to DLBCL transformation provides an excellent tool

to examine the response of the Myc signature genes because the transformation from the

less aggressive follicular lymphoma to the more aggressive diffuse large B-cell

lymphoma has been correlated in many cases with mutations that lead to deregulation of

MYC. Indeed, the Myc signature gene expression profile of DLBCL samples display a

greater resemblance to the Myc signature in fibroblast than do the FL patient samples.

This observation is consistent with the hypothesis that Myc deregulation underlies the FL

to DLBCL transformation. However, when we look at individual genes, the FL to

DLBCL expression profile does not uniformly approach the fundamental Myc signature

response. Expression of some genes is either not different between FL and DLBCL or is

opposite to what would be predicted (assuming a MYC event has taken place) by the

Myc signature. There are at least two related possible reasons for these inconsistently

responding genes. First, it is important to note that the study looked at the transformation

of one malignancy to another, rather than the transformation from a normal state to a

malignant state. It is therefore possible that some of the Myc signature genes are under

control of other abnormally expressed genes and differently influenced by MYC

deregulation. Second, due to possible cell type specific regulation of Myc signature

genes, some genes may be refractory to, or differently influenced by, MYC deregulation.

75

The Myc signature centroid serves as the fundamental Myc signature response

to which other expression profiles are compared. The correlation between the Myc

signature and an individual expression profile quantitatively describes the similarity. We

were interested in seeing if an increased resemblance to the Myc signature correlated

with increased c-Myc RNA levels. In tumor expression panels, the c-Myc RNA levels

(as measured on the microarray) correlates with the presence of the Myc signature.

Normal human tissues, however showed a discordance between the Myc signature and

the c-Myc RNA levels (Figure 2.22). We were surprised not only at the discordance, but

also by the fact that highly proliferative tissues from the gastro-intestinal tract did not

display the Myc signature. One explanation for this finding is that the Myc signature

defined in this study may describe a signature profile in cells which have experienced

deregulated and supra-physiological levels of Myc. Finding that c-Myc RNA levels more

closely match the presence of the Myc signature in tumor panels supports the notion that

the Myc signature defined here is particularly useful in identifying a “deregulated Myc

signature” in clinical expression profiles.

As previously mentioned, correlations between Myc aberrance and survival are

conflicting. Using a relatively simple starting metric such as “high Myc levels” to

correlate with a complex end metric such as “survival” may be prone to inconsistencies

The Myc signature, however, may prove to be a multifaceted starting metric that is far

more reliable in predicting survival. As is evident when we clustered tumors datasets,

even though those tumors that bear the Myc signature generally have the highest relative

Myc levels, there are tumors among these subclusters that are relatively low in Myc yet

still display the Myc signature. Furthermore, there are tumors that have relatively high

76

Myc levels, but do not display the Myc signature. Because our approach identifies a

broad Myc response rather than simply the variation in expression of one gene (i.e.

MYC), this methodology has the ability to classify tumors with a more defined

transcriptional program in progress, and the knowledge of which may have implications

for clinical decisions. Looking forward, certain therapies may be discovered to be

particularly effective or ineffective in treating patients of cancer that display the Myc

signature. Therefore, the methodology we have described of identifying the Myc

signature in malignant tissue may be critical for accurate diagnosis of cancers that bear

the Myc signature and clinically useful in deciding therapeutic strategies.

After clearly defining signature responses or transcriptional modules to given

stimuli, these responses can not only be used to query expression profiles of clinical

samples, but these modules can be directly compared to each other. For example, we

compared the transcriptional response in fibroblasts to MYC overexpression and serum

stimulation. This is of particular relevance because MYC is an immediate early serum

response gene (MYC expression is induced by serum stimulation even in the absence of

protein synthesis). When the Myc signature is examined in the serum response, nearly all

the Myc signature genes are also serum responsive. Interestingly, however, the direction

of response to Myc and serum is not always the same. There are at least two possible

mechanisms that may explain this observation. First, serum is a heterogeneous stimulus

that is likely to activate many different pathways. The final expression levels of some

genes may be the consequence of dueling regulatory influences of different pathways.

Hence, while serum stimulation induces MYC expression, which may alone lead to the

repression of a gene, other serum activated pathways may lead to a positive regulatory

77

influence on that same gene. In the end, the direction of regulation may be the

consequence of who “wins” this regulatory battle. Second, deregulated Myc expression

may have different and possibly opposing effects on Myc target genes. The

transcriptional effects of serum on Myc signature genes may reflect the response to

normally regulated Myc, while the transcriptional effects of Myc overexpression may

reflect the response to deregulated Myc expression.

Genes repressed in response to Myc expression are dominated by genes involved

in both recognizing signals from and sending signals to the external milieu. This raises

the possibility that cells overexpressing Myc are less responsive to external signals. In an

experiment that measured responsiveness to FBS, cells expressing activated MycER were

indeed less responsive at the level of ERK 1/2 activation. The cells’ inability to respond

to growth stimulatory cues such as growth factors in FBS may be the result of Myc

initiated negative feedback loops of cell signaling genes. This reduced ability to respond

to external cues may underlie Myc’s ability to inhibit differentiation, because Myc

overexpressing cells may also be refractory to other environmental cues which are

differentiative. Indeed, many tumor cells exist in tissue microcosms that would otherwise

encourage differentiation of cells to a terminally differentiated fate specific to that tissue.

Hence, the ability to not respond to such signals may be a critical quality of oncogenic

transformation. Reduced ability to respond to all external signals is not a foregone

conclusion, however, because some of the communication genes, such as DKK1 and

DKK3, are actually inhibitors of signaling in the canonical WNT pathway; the repression

of signaling inhibitors may actually lower a threshold to respond to cues. Intriguingly,

78

Wnt pathway members, Wnt5a and Wnt5b, are similarly repressed, and, they are

agonists of the noncanonical WNT pathway.

On the flip side of growing independent or unresponsive to tissue specific signals,

oncogenic transformation may be predicated on the cells’ ability to not be recognized by

surrounding tissues as ‘rogue’ cells. In this theme of aberrant cells keeping a low profile,

genes such as HLA-A, -E, -F and –G that are involved in class I antigen presentation are

repressed by steady state Myc overexpression. While all of these genes show some

repression by MycER, HLA-G is most convincingly repressed by MycER activation.

Therefore, we speculate that the class I antigen presentation pathway may be

downregulated and that Myc expressing cells may be less likely to stimulate an immune

response to tumor specific antigens (such as viral proteins or those that arise from

translocations or other mutations) that would otherwise be presented. Further

experiments are necessary to test this hypothesis.

In conclusion, we have defined a fundamental transcriptional response to MYC

overexpression in an in vitro cell system of mesenchymal origin. Knowledge of the Myc

signature guided our query and facilitated the identification of active Myc programs in

complex expression profiling datasets of clinically relevant and non-mesenchymal origin.

The transcriptional response to deregulated MYC expression, as defined in our study,

more closely resembles a Myc initiated transcriptional program in malignant tissues than

in normal human tissues. Given that the Myc signature was defined in a system in which

supraphysiological levels of Myc are generated in a cell type that already expressed

normal MYC levels, the identified responsive genes are those that are capable of being

further regulated by this higher dose of Myc. Hence, this experimental system to define

79

the Myc signature closely resembles a putative malignant process driven by MYC

deregulation. Further investigation of this deregulated Myc signature may provide

potential drug targets whose specific modulation would have few consequences on

normal cellular proliferation. Finally, the ability to effectively identify underlying drivers

of malignant transformation may one day have substantial impact on clinical therapeutic

decisions.

80

Materials and Methods

Cell Culture, Transfection and Retroviral Infection

Primary human foreskin fibroblasts (BJ), retroviral producer PhoeNX cells were cultured in

DMEM supplemented with 10% fetal bovine serum. Retroviral infection of BJ cells was

performed using PhoeNX cells. PhoeNX cells were transfected via the calcium phosphate

method. To obtain a transduced population, after completion of infection cells were

selected for resistance to hygromycin (150 µg/ml) or puromycin (0.8 µg/ml) for 7 days.

Western blotting

Protein lysates were resolved by standard SDS-PAGE methods. Proteins were then

transferred to PVDF using a wet transfer apparatus. Membranes were then blocked for 10’

to O/N in TBS (+0.1% tween 20 and 1% skim milk). FLAG western blots were probed

with diluted (1:5000) FLAG antibody (M2, Sigma) in blocking solution. Blots were

washed 3 times with TBS (+0.1% tween 20). Diluted (1:10,000) secondary antibody (HRP

conjugated goat anti-mouse) was used to detect primary FLAG antibody. Enhanced

chemoluminescence reagents (Amersham Biosciences) were used to detect HRP. Kodak

XAR film was exposed to treated membranes. Similar procedures were used for anti-

ERK1/2 antibodies except for some manufacturer (Cell signaling technologies, MA)

directed changes in blocking times and blocking buffer.

Reverse Transcription and PCR

Total RNA was isolated from infected cells using TRIzol reagent (GIBCO-BRL). Reverse

transcription of 1 µg of RNA was performed using SuperScriptTM First-Strand Synthesis

System (GIBCO-BRL), according to the manufacturer instructions. PCR was performed

81

on cDNA using the following sets of oligos: human TERT-specific oligos: HT1

(CGGAAGAGTGTCTGGAGCAA) and HT2 (GGATGAAGCGGAGTCTGGA); human

GAPDH-specific oligos: HG1 (CTCAG-ACACCATGGGGAAGGTGA) and HG2

(ATGATCTTGAGGCTGTTGTCATA); human RUVBL1-specific oligos: SC133

(GGAGGTGAAGAGCACTACGAA) and SC134 (GCCAACAAGACAGCTCTTCC);

human MAP3K10-specific oligos: SC135 (TACACAATGATGCCCCTGTG) AND

SC136 (TCATCTTGGTGGTCTTGTGC); human EBNA1BP2-specific oligos: SC137

(CTGAAGCAATGTTTGGCAGA) and SC138 (CGCTTCGTAGGGACTTTGAG);

human MYL9-specific oligos: SC139 (CATCCAATGTCTTCGCAATG) and SC140

(CATCATGCCCTCCAGGTATT); UOligos specific for human and rat CAD cDNA:

CAD1 (GGTGGATCTGGAG-CATGAGTGGA) and CAD2

(AGATGGAAGCGGCCATCAGGAAG). For quantitative analysis of PCR products

oligos HT1, HG1, CAD1 were end labeled with γ32P-ATP. Amplification was performed

in a T3 Thermocycler (Biometra®). PCR products were resolved on 6% polyacrylamide

gel and quantitated using Molecular Dynamics Phosphor Imaging System.

RNA preparation, labeling, and microarray hybridization

Total RNA was prepared from cells at times and conditions indicated in text using

the TRIzol reagent (GIBCO-BRL) according to the manufacturer’s protocol. Reference

RNAs were made from rapidly dividing untransduced BJ cells. RNAs were quantitated

using a spectrophotometer. The Agilent low RNA input fluorescent linear amplification

kit was used to generate labeled cRNA for time course experiments. 0.5 µg of total input

RNA was used for time course experiments. The Agilent fluorescent direct label kit was

used to generate labeled cDNA for steady state experiments. 10 µg of total input RNA

82

was used for steady state experiments. Manufacturer protocols were followed. Cy3-

CTP and Cy-5 CTP were obtained from Perkin Elmer/NEN Life Sciences. Agilent

human 1A version 2 arrays were used in both the time course and steady state studies

except for the five vector control arrays in the time course study; those five experiments

were performed with agilent human 1A version 1 arrays. Hybridizations and washes

were performed according to the manufacturer’s protocol. Washed arrays were scanned

using the Axon Intruments GenePix 4000B microarray scanner and intensities collected

by using GenePix5.0 software. All data were stored in the UNC Microarray Database.

Microarray data analysis

Flagged spots for “spot not found” and “spot bad” were removed from data. The

background subtracted lowess (for time course data) or global linearly (steady state

experiment) normalized data were subject to a filter that demanded a spot intensity in

either channel was greater than 1.2-fold above background intensity. In each time course

series (LCPX, MycER, or Myc(∆MBII)ER), data for at least two of the three zeroes (for a

given probe) must have met the above criteria for the whole series to pass the next filter.

Similarly for the steady state experiments, data for both vector arrays (for a given probe)

must have met the above criteria for the duplicate Myc and Myc∆MBII arrays to pass this

filter. Subsequently, only those probes that met the above criteria for 80% of the 25

arrays in the time course experiment or of the 6 arrays in the steady state experiment. For

a given probe in a time course series, the average expression value across the three zeros

was subtracted from every expression value in that series. This step is called “zero-

transformation.” A similar zeroing was performed on the steady state data except the the

vector expression values served as the zeroes. Time course data were subject to a 1.7

83

fold in at least two arrays filter. The probes (1631 probes) that pass these filters for the

time course experiment are considered MycER responsive. SAM (Significance Analysis

of Microarrays, version 1.21) was used to perform a two class, unpaired analysis of the

steady state data with the wild type Myc arrays in one class and the vector and mutant

Myc arrays in another. The resulting probes (1608 probes) from this analysis were

considered Myc responsive.

Two clustering algorithms were used in this study: hierarchical and k-means.

Hierarchical clustering analyses were performed with Cluster 3.0 and k-means clustering

analyses were performed in the MeV from TIGR.

The probes defined as both MycER and Myc responsive were considered the Myc

signature genes. The expression pattern of these genes was defined as the Myc gene

expression signature. The Myc signature centroid was calculated in three steps. First, the

average expression values for the Myc signature genes across the last three time points

(24 HRs, 36 HRS, 48 HRS) in the time course experiment were determined. Second, the

average expression values for the Myc signature genes across the duplicate Myc arrays

were determined. Finally, the average of these two averages resulted in the Myc centroid.

When different expression profiling datasets were linked together, publicly

available datasets were first filtered, normalized and centered as in the original study.

Then, platform-specific clone identifiers were converted (using Source and other utilities)

to unigene cluster identifiers (unigene build: March 2005). Data from multiple spots

mapping to the same unigene cluster ID were averaged across the multiple spots. The

data were then linked via unique unigene cluster ID’s. In all data presented as Java

Treeview images, the contrast setting was set to “2”.

84

Immunohistochemistry

Paraffin-embedded breast tumors were cut into 5 mm sections. Tissue sections were

deparaffined with xylene, dehydrated with ethanol and endogenous peroxidase activity

was blocked with a 3% hydrogen peroxide solution. The slides were incubated with

10mM citrate buffer (pH 6.0) and microwaved for 20 min for antigen retrieval. The slides

were then blocked with goat serum and incubated with cMYC antibody for 30 min (Santa

Cruz sc-40, 1 : 50 dilution), and then incubated with biotin conjugated goat anti-mouse

IgG (Vector Laboratories, CA, USA). Proteins were visualized with streptavidin-

conjugated HRP (Vector Laboratories, CA, USA). The slides were counterstained with

50% hematoxylin and examined by light microscopy on a Zeiss microscope at x100

magnification. These experiments were performed in the Perou laboratory.

85

CHAPTER 3: A Functional Screen for Myc-responsive Genes Reveals

SHMT: a Major Source of One-carbon Unit for Cell Metabolism

The work presented in the following chapter is published. The reference for this paper is the following and is available as a supplemental file: Nikiforov MA, Chandriani S, Petrenko O, O’Connell B, Kotenko I, Beavis A, Sedivy JM, Cole MD 2002. Functional screening for Myc-induced genes reveal serine hydroxymethyltransferase, a major source of one-carbon unit for cell metabolism. Molecular and Cellular Biology 22: 5793-5800.

86

Results

Functional screening for Myc-responsive genes

The creation of c-myc-null cells through targeted knockout of both somatic genes in a rat

fibroblast cell line has contributed significantly to an understanding of Myc function

(Bush et al. 1998) (Mateyak et al. 1997) (Xiao et al. 1998). These cells offer an

opportunity to assess Myc-dependent phenotypes without manipulation of the cellular

environment and to assay the contribution of individual potential target genes to growth

control. Attempts to complement c-myc-null cells with cDNA expression libraries

resulted in the finding that only c- and N-myc cDNAs would suffice, implying a unique

function for Myc in cell proliferation (Berns et al. 2000) (Nikiforov et al. 2000). These

data also suggested that Myc fulfills its function through the regulation of more than one

downstream effector. Thus, in the course of the selection, c-myc-null cells expressing c-

or N-myc cDNA will most likely outgrow cells in which expression of a cDNA of a Myc-

target gene only partially complements the cellular growth defects. Since at least one of

the myc family genes is expressed in every cell type, all previously available cDNA

libraries invariably contain myc cDNAs. To overcome this obstacle, we prepared cDNA

from N-myc reconstituted c-myc-null cells and depleted it of full length N-myc cDNA

(Figure 3.1 and Materials and Methods), thus creating a library enriched with cDNAs of

N-Myc target genes yet lacking functional N-myc cDNA. The Myc-depleted cDNA

library was introduced into DK cells, a c-myc-null cell line which overexpresses a cdk4-

cyclinD1 fusion protein (Rao et al. 1999), which partially complements their growth

defects (Figure 3.2D). To select for the faster growing cells, we utilized a CFSE/FACS

sorting methodology described earlier (Nikiforov et al. 2000). Briefly, cells are pulse-

87

Sfi ITransduction of

N-myc cDNA

RNA IsolationConstruction of a cDNA Library

Digestion with Sfi I

N-myc

Library Transduction

N-myc

CFSE-based Cell Sorting

Sort 1 Sort 2

HO15.19 cells

DK cells

CFSE

Cel

l Num

ber

Sfi ISfi ITransduction of

N-myc cDNA

RNA IsolationConstruction of a cDNA Library

Digestion with Sfi I

N-myc

Library Transduction

N-myc

CFSE-based Cell Sorting

Sort 1 Sort 2

HO15.19 cells

DK cells

CFSE

Cel

l Num

ber

Figure 3.1 Outline of the experimental strategyDotted line indicates CFSE profile of the initial population before the first cell sorting. Shadowed area below each peak designates cells taken for the next step of analysis. Note a gradual decrease in the intensity of CFSE staining of cells after each sorting. See details in the text and Materials and Methods.

88

labeled with the irreversibly binding vital fluorescent dye carboxyfluorescein diacetate

succinilmidyl ester (CFSE) and propagated for five days in culture to allow the dye to be

diluted through cell division. Cells were then subjected to FACS sorting based on the

intensity of staining. In five days, faster-growing cells will retain lower amounts of CFSE

compared to slow-growing cells, since the number of cell divisions corresponds to a CFSE

dilution factor. After two rounds of sorting, a population of faster-growing cells was

obtained, although the rate of their growth was lower than that of Myc-reconstituted c-

myc-null cells (Figure 3.2A). Retroviral inserts were rescued from the faster growing

population (Materials and Methods) and subjected to sequence analysis. Isolated cDNAs

with intact coding regions include proteosome component C3, Fos-related antigen 1 (Fra1),

ribosomal Protein S28, vacuolar H+ ATPase subunit E, ribosomal protein S20, and

mitochondrial serine hydroxymethyltransferase (mSHMT). We also isolated several

truncated cDNAs for α2(I)collagen.

mSHMT partially rescues the slow growth phenotype of c-myc-null cells

cDNAs identified in the course of the screening were individually transduced into DK

cells in parallel with an empty vector control, and subjected to the original selection. Cells

with low levels of CFSE staining were collected using a cell sorter and a growth rate of

the sorted cells was determined using FACS analysis. Among all tested cDNAs,

overexpression of only mitochondrial serine hydroxymethyltransferase (mSHMT) resulted

in a low but reproducible increase in the growth rate of DK cells compared to that of cells

transduced with vector only (Figure 3.2B). Mitochondrial SHMT and a different gene

encoding its cytoplasmic isoform (cSHMT) are both nuclear genes. SHMT enzymes

catalyze the conversion of serine and tetrahydrofolate (THF) into glycine and 5,10-

89

methylene-THF. This reaction is a major source of one-carbon-substituted folate

cofactors necessary for a variety of metabolic reactions, including thymidylate and de

novo purine biosynthesis, biosynthesis of methionine and histidine, and the catabolism of

methyl groups and formate (reviewed in (Snell et al. 2000). It has also been suggested that

SHMT activity may regulate cell growth and proliferation (Snell 1985) (Snell 1980)

(Thorndike et al. 1979).

We were interested in whether overexpression of mSHMT rescues the slow growth

of the original c-myc-null cells (HO15.19). To this end, we transduced a mSHMT-

expressing vector or an empty vector into c-myc-null cells or DK cells, followed by

selection on hygromycin for 7 days. The growth rate of the resulting cell populations was

determined by both CFSE staining and the direct determination of the cell doubling time

(Figure 3.2). Overexpression of mSHMT partially rescues the slow growth phenotype of

HO15.19 and DK cells as visualized by the peak of staining with CFSE (Figure 3.2C).

The doubling time of the c-myc-null cells decreased from 47.1± 1.9 hours to 39.0±1.4

hours for myc-null/mSHMT cells (n=4, p=0.0173), and from 40.9± 1.8 hours for DK cells

to 35.5±1.3 hours for DK/mSHMT cells (n=3, p=0.0333). It is noteworthy that

overexpression of mSHMT in the parental Rat1 cells does not enhance their growth rate

(Figure 3.2D).

Mitochondrial and cytosolic SHMT genes are direct targets of Myc

Our next goal was to determine if mSHMT is indeed a Myc-target gene. We used a

Northern blot analysis of RNA isolated from c-myc-null cells and parental Rat-1 cells

(which possess wildtype c-myc genes). As shown in Figure 3.3A, mSHMT expression is

strongly induced by serum in Rat-1 cells, whereas the induction is severely attenuated in

90

100 101 102 103 104

0

10

20

30

40

50

60

vector

mSHMTc-myc

vector

mSHMTvector

mSHMT

HO15.19 DK Rat1

Dou

blin

g tim

e (h

ours

)

B

100 101 102 103 104

C

D

Cel

l num

ber

(line

ar)

100 101 102 103 104

A

CFSE100 101 102 103 104100 101 102 103 104100 101 102 103 104

0

10

20

30

40

50

60

vector

mSHMTc-myc

vector

mSHMTvector

mSHMT

HO15.19 DK Rat1

Dou

blin

g tim

e (h

ours

)

B

100 101 102 103 104

C

100 101 102 103 104100 101 102 103 104100 101 102 103 104

C

D

Cel

l num

ber

(line

ar)

100 101 102 103 104100 101 102 103 104100 101 102 103 104

A

CFSE

Figure 3.2 Overexpression of mSHMT enhances proliferation of c-myc-null cellsA. FACS profile of DK cells (thin line), c-myc-null cells (HO15.19) infected with a c-myc-expressing vector (dotted line), and DK cells infected with a cDNA-expression library after two rounds of sorting (bold line). B. FACS profile of DK cells infected with either empty vector (thin line) or a mSHMT-expressing vector (dashed line) after one round of sorting. c-myc-null cells (HO15.19) infected with a c-myc-expression vector before the sorting are shown with a dotted line. C. CFSE profile of c-myc-null cells (HO15.19) infected as in B and selected for resistance to hygromycin (see Materials and Methods). D. Doubling time of DK, c-myc-null, and Rat1 cells expressing the constructs indicated below each bar.

91

c-myc-null cells. We also compared the profile of mSHMT expression in logarithmically

growing Rat-1 cells, c-myc-null cells and c-myc-null cells reconstituted with a c-myc

cDNA. The level of mSHMT RNA was dramatically enhanced in Myc-reconstituted c-

myc-null cells compared to the original c-myc-null cells transduced with empty vector,

implying that mSHMT is a downstream target of c-Myc. We also noted a second, slower

migrating mSHMT RNA in serum stimulated myc-null cells that is currently under

investigation. We note that mSHMT is still expressed basally in log phase myc-null cells

and inducible to some extent by serum, indicating that other factors also regulate mSHMT

expression.

To further prove that mSHMT regulation was Myc-dependent, we employed a Myc-

ER inducible system (Littlewood et al. 1995). Myc-null cells were infected with a vector

that expresses a chimeric protein consisting of c-Myc fused to the hormone-binding

domain of the estrogen receptor. Under normal conditions, the c-Myc-ER chimera is not

transported to the nucleus and thus does not activate Myc-responsive genes. Upon

addition of 4-hydroxytamoxifen (4-OH), subcellular translocation regulates c-Myc-ER

activity. We tested the expression level of mSHMT in c-myc-null/Myc-ER cells treated or

untreated with 4-OH. 4-OH induces a substantial up-regulation of mSHMT expression,

whereas there was no induction of mSHMT in untreated cells (Figure 3.3B) or in cells

expressing vector only with or without 4-OH (data not shown). Induction of Myc-ER by

4-OH in these cells fully restores their growth rate to wild type levels (data not shown).

mSHMT is also induced 2.4-fold by Myc-ER in the presence of cycloheximide (data not

shown). As mentioned above, there are two serine hydroxymethyltransferases

(mitochondrial and cytoplasmic) that are major suppliers of one-carbon units for cellular

92

A

C D

HCT116

COLO 32

0SK-N

SHSK-N

ASCHP 13

4IM

R 32

mSHMT

GAPDH

cSHMT

B0h 4h 8h 12h 16h

- + - + - + - + - +4-OHtime (h)

mSHMT

GAPDH

1.0:1.0 1.0:0.9 1.0:1.9 1.0:2.6 1.0:3.3 1.0:4.0

- +48h

0h 1h 3h 6h 12h 24hCycling cells

+ - + - + - + - + - + - + -c-myc

Serum stimulation

rRNA

mSHMT

m -

mSHMT

cSHMT

N-mycc-myc

rRNA

m+ -

+/+-/-

G1 S G2/MG0G1G0 S

A

C D

HCT116

COLO 32

0SK-N

SHSK-N

ASCHP 13

4IM

R 32

mSHMT

GAPDH

cSHMT

B0h 4h 8h 12h 16h

- + - + - + - + - +4-OHtime (h)

mSHMT

GAPDH

1.0:1.0 1.0:0.9 1.0:1.9 1.0:2.6 1.0:3.3 1.0:4.0

- +48h

0h 1h 3h 6h 12h 24hCycling cells

+ - + - + - + - + - + - + -c-myc

Serum stimulation

rRNA

mSHMT

m -

mSHMT

cSHMT

N-mycc-myc

rRNA

m+ -

+/+-/-

G1 S G2/MG0G1G0 S

Figure 3.3 mSHMT and cSHMT are Myc-responsive genes

93

Figure 3.3 mSHMT and cSHMT are Myc-responsive genesA. RNA isolated from serum stimulated or logarithmically growing c-myc-nullcells (-), parental Rat1 cells (+), or c-myc-null cells expressing a c-myc cDNA (m) was probed in Northern blotting with a mSHMT-specific probe. 28S and 18S ethidiumbromide-stained rRNAs are shown in each lane for a loading control. The numbers on the top indicate time after serum induction. B. c-myc-null cells expressing c-MycER were passaged several times under subconfluent conditions (<50%) to obtain cells in an exponential phase of growth. 4-hydroxytamoxifen (4-OH) was added directly to the media at a final concentration of 200 nM. RNA was collected at the time intervals post-4-OH addition shown on the top (in hours). RNA was also collected during the same time intervals from untreated cells. RNA was probed by Northern blot sequentially with mSHMT and GAPDH probes. The numbers below indicate the fold induction of mSHMT-specific message normalized to the amount of GAPDH in each lane for each time point. C. RNA isolated from logarithmically growing c-myc-null cells (-), parental Rat1 cells (+), or c-myc-null cells expressing a c-myc cDNA (m) was sequentially hybridized in Northern blotting with probes specific for mSHMT, cSHMT, and GAPDH. D. RNA was isolated from logarithmically growing colon carcinoma cells (HCT116 and Colo320) or neuroblastoma cells (SK-NSH, SK-NAS, CHP-134, IMR32) and sequentially hybridized in Northern blotting to probes specific to mSHMT, cSHMT, c-myc or N-myc genes. 28S and 18S ethidium stained rRNA is shown in each lane for a loading control.

94

biosynthesis. Since we demonstrated that mSHMT is a Myc-target gene, we were

interested to determine if expression of cSHMT is also Myc-regulated. To this end, we

hybridized RNA obtained from Rat-1 cells, c-myc-null cells and Myc-reconstituted c-myc-

null cells with a probe specific to mouse cSHMT in Northern blotting. As shown in

Figure 3.3C, the level of cSHMT mRNA is significantly higher in cycling Rat1 or Myc-

reconstituted c-myc-null cells than in cycling c-myc-null cells infected with empty vector.

Expression of cSHMT was also induced in Myc-ER cells by treatment with 4-

hydroxytamoxifen, suggesting that cSHMT is also a Myc-target gene (data not shown).

Since c-myc and N-myc genes are often amplified in tumors, we were interested if

the level of mSHMT or cSHMT mRNA is also elevated in tumor cells with high amounts

of the Myc protein. SHMT expression was analyzed in colon carcinoma cell lines with

amplified (Colo320) or normal (HCT116) c-myc genes and in neuroblastoma lines with

amplified (CHP134, IMR32) or normal (SK-NSH, SK-NAS) N-myc genes. As shown in

Figure 3.3D, expression of both SHMT genes is higher in tumor cell lines with amplified

myc genes than in cells with lower Myc levels.

In vivo crosslinking of c-Myc to the SHMT promoters.

Sequence analysis of cytosolic and mitochondrial SHMT genes in different species

revealed that their promoters contained one or two Myc consensus binding sites (Figure

3.4A). We wanted to determine if Myc actually binds to these sites in vivo. To this end,

we obtained cross-linked chromatin from either myc-null cells and Myc-reconstituted null

cells (Figure 3.4B) or from human cells HEK293 and Raji, which have high levels of the

c-Myc protein (Figure 3.4C). The chromatin was immunoprecipitated with either anti-c-

Myc antibodies or with control anti-Gal4 antibodies (ChIP). After reversion of the cross-

95

linking, DNA was purified and subjected to PCR with primers flanking the region of the

SHMT promoters encompassing the potential Myc binding sites (Figure 3.4A). As a

control for DNA quantitation and nonspecific background, we performed PCR on the

same DNA samples with primers to chromosomal sites that are not expected to bind to

Myc protein in vivo. In myc-null cells, there was no enrichment in the anti-Myc ChIP for

the Myc target gene nucleolin or for mSHMT, nor was there any enrichment for the

nontarget genes glucokinase (which contains a Myc/Max consensus site) and PCNA. In

contrast, reconstitution of Myc expression in myc-null cells led to 4-5-fold enrichment of

both mSHMT and nucleolin genomic DNA fragments in the anti-Myc ChIP but not in the

control anti-Gal4 ChIP samples (Figure 3.4B). No binding of Myc to glucokinase or

PCNA was observed. In Raji and HEK293 cells, there was a substantial enrichment for

mSHMT and cSHMT genomic DNA in the anti-Myc ChIP (3.5-20-fold) that paralleled

the enrichment for the Myc target gene TERT (2.4-11-fold) (Figure 3.4C). There was no

enrichment of a nonfunctional Myc consensus sequence on chromosome 22 or for the

nontarget ß-globin gene. Equivalent background signals were obtained for all samples

with the control ß-globin primers. It is noteworthy that the highest signal-to-noise ratio

was observed for the mSHMT gene in Raji anti-Myc ChIP samples (20 fold) which have

the highest Myc protein levels. Thus, chromatin immunoprecipitation demonstrates the

direct binding of c-Myc to both the mSHMT and cSHMT promoters, consistent with the

data above on Myc-dependent expression of these genes.

Discussion Several approaches have been used to discover genes whose expression is modulated by

Myc in normal or transformed cells. This study represents the first successful functional

96

A

tCACGTGg gCACGTGc

mouse cSHMTATG

human cSHMT

tCACGTGg

ATG

gCACGTGcgCACGTGt

rat mSHMTATG

gCACGTGg tCACGTGg

human mSHMTATG

B C

mSHMT

NUCL

PCNA

GLU

1.0 3.5 1.0 8.7

1.0 4.5 1.0 20.7

1.0 1.0 1.0 1.3

Raji 293 G GM M

mSHMT

cSHMT

TERT

Chr. 22

β-globin

1.0 2.4 1.0 11.8

1.0 0.8 1.0 5.3

1.0 1.4 1.0 4.5

1.0 1.2 1.0 0.9

+myc+vector G GM M

HO15.19

A

tCACGTGg gCACGTGc

mouse cSHMTATG

tCACGTGg gCACGTGc

mouse cSHMTATGATG

human cSHMT

tCACGTGg

ATGhuman cSHMT

tCACGTGg

ATG

gCACGTGcgCACGTGt

rat mSHMTATG

gCACGTGcgCACGTGt

rat mSHMTATGATG

gCACGTGg tCACGTGg

human mSHMTATG

gCACGTGg tCACGTGg

human mSHMTATG

B C

mSHMT

NUCL

PCNA

GLU

1.0 3.5 1.0 8.7

1.0 4.5 1.0 20.7 1.0 4.5 1.0 20.7

1.0 1.0 1.0 1.31.0 1.0 1.0 1.3

Raji 293 G GM M

mSHMT

cSHMT

TERT

Chr. 22

β-globin

1.0 2.4 1.0 11.81.0 2.4 1.0 11.8

1.0 0.8 1.0 5.3

1.0 1.4 1.0 4.5

1.0 1.2 1.0 0.9

+myc+vector G GM M

HO15.19

Figure 3.4 c-Myc binds to the promoters of mSHMT and cSHMT genes in vivo

97

Figure 3.4 c-Myc binds to the promoters of mSHMT and cSHMT genes in vivoA. Schematic presentation of human, mouse and rat mSHMT and cSHMT promoters. Open boxes represent exons. Myc-binding sites and ATG codons are shown. Arrows indicate the position of primers used in PCR. B-C. HO15.19 cells infected with empty vector (vector), HO15.19 cells infected with c-myc-expression vector (myc), HEK 293 cells or Raji cells were cross-linked, lysed, and chromatin was immunoprecipitated with c-Myc-specific (M) or Gal4-specific (G) antibodies followed by the reversal of the cross-linking and DNA isolation. Isolated DNA was used in PCR with radiolabeledoligos flanking Myc binding sites in the promoter of the genes indicated on the left side. For negative controls, two types of PCR were performed. The first one was carried out with oligos flanking "CACGTG" sites that do not interact with c-Myc in the promoter of rat glucokinase gene (GLU) (11) (B) or somewhere on human chromosome 22 (4)(C).Another PCR was performed with oligos flanking a DNA region containing no "CACGTG" sites in the rat PCNA gene (B) or in the third intron of a silent human ß-globin gene (C). Quantitation of the PCR products was performed using the Molecular Dynamics Phoshphor Imaging System. Numbers below a panel indicate the fold difference in the signal intensity determined by dividing a given PCR signal by a corresponding PCNA-specific (B) or �-globin-specific signal (C) and by the ratio of these signals in the "Gal4" lane.

98

screen for cDNAs that can bypass Myc function. Two previous attempts to complement

the growth defect in myc-null cells led to the repeated cloning of Myc itself: either c-Myc

or N-Myc (Berns et al. 2000); (Nikiforov et al. 2000). These studies emphasized the

unique nature of the Myc pathway in cellular physiology but offered little insight into the

functional downstream target genes. Here we introduce a novel functional screen for

downstream effectors of Myc function, using a powerful sorting technique for cellular

growth rate and a library depleted of myc cDNAs. We show that an important aspect of

Myc function is to regulate cellular one-carbon metabolism through the transcriptional

regulation of genes encoding both mitochondrial and cytoplasmic SHMT. Neither of

these genes have been previously identified as Myc targets using subtractive or differential

hybridization. Both genes have functional Myc/Max binding sites near or within their

TATA-less promoters, and both genes can be cross-linked to Myc in vivo. SHMT genes

are Myc-dependent in steady-state log phase growth in rat fibroblasts and induced by

Myc-ER in 4-OH-regulated MycER reconstituted cells. These data establish that

mSHMT and cSHMT are direct targets of Myc in human and rodent cells and that Myc is

a rate-limiting factor for their expression.

SHMT catalyzes the conversion of serine and tetrahydrofolate (THF) to glycine and

methylene-THF (reviewed in (Snell et al. 2000). This reaction generates most of the one-

carbon unit for thymidylate, purine, and methionine synthesis. Two separate genes encode

the mitochondrial (mSHMT) and cytoplasmic (cSHMT) forms of the enzyme (Garrow et

al. 1993) (Stover et al. 1997; Girgis et al. 1998). It remains unclear as to why a cell has

both enzymes, since the mitochondrial form is the main source of one-carbon units for

both mitochondrial and cytoplasmic folate metabolism and nucleoside biosynthesis (Fu et

99

al. 2001). Interestingly, culturing of myc-null cells or myc-null cells overexpressing

cyclinD1/cdk4 in high concentrations of nucleotides, formate, or methionine does not

provide any growth enhancement (data not shown), indicating that it is not possible to

complement this biochemical pathway by simple substrate replacement. This suggests

that localized SHMT activity produces one-carbon-substituted folate cofactors that may

have a unique influence on cell growth and/or proliferation. In yeast, the two SHMT

enzymes are not essential due to redundant biosynthetic pathways. However, strains

defective in both enzymes exhibit severe growth defects under specific culture conditions

(Kastanos et al. 1997). Furthermore, CHO cells lacking mSHMT are glycine auxotrophs

(Pfendner and Pizer 1980). We speculate that mSHMT activity may become rate-limiting

for growth in c-myc-null cells in part because of other metabolic pathways that are

simultaneously downregulated in response to the Myc deficiency.

The full extent of the metabolic pathways that are under Myc control will require

further study. The SHMT genes join CAD, ODC, LDH-A, IRP2 and others as direct Myc

targets involved in basic cellular metabolism (Bello-Fernandez et al. 1993) (Miltenberger

et al. 1995) (Shim et al. 1997) (Wu et al. 1999). Each of the latter genes are components

of general cellular growth pathways, but they have not been shown to contribute directly

to growth rate as we show here for mSHMT. The importance of Myc in the control of

cellular growth is highlighted by studies in Drosophila, where Myc gain-of-function or

loss-of-function controls cell size without any concomitant change in cell number

(Johnston et al. 1999). These observations link Myc expression to cellular growth and

only indirectly to the cell cycle. Overexpression of Myc in mouse B cells also leads to an

increase in cell size (Iritani and Eisenman 1999). In contrast to these influences on

100

cellular growth, several mammalian genes with direct roles in control of the cell cycle

have been shown to be Myc targets, including cyclinD2, cdk4, and cul1. Cul1 has been

reported to rescue the growth defect in primary myc-null fibroblasts (O'Hagan et al. 2000),

and cdk4 has a modest growth enhancing effect in immortalized c-myc-null fibroblasts

(Hermeking et al. 2000). CyclinD2 is reported to be required for Myc-dependent cell

proliferation (Bouchard et al. 1999), but ectopic expression of cyclinD2 does not stimulate

the growth of c-myc-null cells (Berns et al. 2000). Finally, reduced expression of c-myc

led to smaller embryos with fewer cells in most major organs, rather than smaller cells

(Trumpp et al. 2001). Thus, it appears that in mammals, Myc controls both basic cellular

growth pathways and progression through the cell cycle. The identification of mSHMT as

a Myc target for cellular growth provides an important insight into Myc’s ability to

regulate cellular metabolism. A screen for cDNAs that synergize with mSHMT to further

enhance the growth rate of Myc-deficient cells may reveal additional rate-limiting

components of the Myc pathway.

101


Cell Culture, Myc-ER System, Transfection and Retroviral Infection.

Rat1 fibroblasts, rat c-myc-null fibroblasts (HO15.19 cells), HO15.19 cells expressing a

cdk4-cyclinD1 fusion protein (DK cells), HO15.19 cells expressing c-MycER chimeric

protein, retroviral producer PhoeNX cells and HEK 293 cells were cultured in DMEM

supplemented with 10% fetal calf serum. Raji cells were cultured in RPMI-1640 medium

supplemented with 10% fetal calf serum. Retroviral infection of HO15.19 and DK cells

was performed according to published protocols (Pear et al. 1993), using PhoeNX cells.

PhoeNX cells were transfected via the calcium phosphate method. To obtain a transduced

cell population, LXSH-based retroviruses HO15.19 cells and DK cells were selected for

resistance to hygromycin (150 µg/ml) for 7 days. HO15.9 cells expressing cMycER

chimeric protein were passaged several times under subconfluent conditions (<50%) to

obtain cells in an exponential phase of growth. 4-hydroxytamoxifen (Sigma, T-176) was

added directly to the media at a final concentration of 200 nM. RNA was collected at 2, 4,

8, and 16 hours post 4-OH addition.

Library Preparation and CFSE Selection

HO 15.19 cells were transduced with pLXSH vector carrying mouse N-myc cDNA, which

contains a unique SfiI site. Digestion with SfiI disrupts the N-myc coding region rendering

it functionally inactive. After infection and hygromycin selection, total cellular RNA was

isolated using TRIzol reagent (GIBCO/BRL) and poly(A) RNA was isolated using

oligo(dT) cellulose (Ambion). cDNA was prepared using SuperScriptTM Choice System

for cDNA Synthesis kit (GIBCO/BRL) and cloned into modified pREBNA vector

102

(Petrenko et al. 1999). The modified pREBNA vector contains bacterial ori sequences

downstream of the polylinker and Lox sequences inserted into the NheI site in the 5' and 3'

LTRs. The library was digested extensively with SfiI, followed by transfection into

packaging cells and subsequent infection into DK cells using standard protocols. After

infection, cells were stained with 10 µM carboxyfluorescein diacetate succinilmidyl ester

(CFSE; Molecular Probes) in suspension in growth media (3x106 cells/ml) for 5 minutes at

370C, washed three times with 10 ml of ice-cold media and plated on 150 mm tissue

culture dish (106 cells/dish). CFSE is a vital fluorescent dye that irreversibly binds to

cellular macromolecules. After five days, cells were collected, analyzed by flow

cytometry and a population of sorted cells (usually 2-5% of cells with the lowest content

of CFSE) was replated. After expanding the cell population on a plate, cells were

subjected to the second round of CFSE-staining and sorting. To rescue retroviral vectors

integrated into cellular genome, we utilized a strategy proposed earlier (Hannon et al.

1999). Briefly, 105 cells after the second sort were infected with a retrovirus encoding the

Cre protein, which binds to Lox sites in LTRs of integrated pREBNA vector and excises a

circular double stranded DNA containing one chimeric LTR, a cDNA insert and bacterial

ori sequences. 36-48 hours post-infection, cells were lysed in buffer P1 from a Qiagen

plasmid mini kit, following treatment with solutions P2 and P3, according to the kit

instruction. Plasmid DNA was cloned through standard transformation of competent E.

coli bacteria.

Chromatin Immunoprecipitation

HEK 293 or Raji cells were cross-linked by adding formaldehyde (Fisher Scientific, final

concentration 1%) directly to the cells on a rocking tissue culture dish for 10 min at room

103

temperature. Cross-linking was terminated by the addition of glycine to a final

concentration of 125 mM. Fixed cells were washed with ice-cold PBS containing 0.5

mM PMSF followed by scraping the cells in swelling buffer (5mM Pipes (pH 8.0),

85mM KCl, 0.5% NP40, 0.5 mM PMSF, and 100 ng/ml leupeptin and aprotinin). Cells

were incubated on ice for 20 min followed by dounce homogenization. Nuclei were

harvested by centrifugation (4500xg) following resuspension in sonication buffer (0.1%

SDS, F-buffer, 0.5 mM PMSF, 100 ng/ml leupeptin and aprotinin) and sonicated on ice to

obtain 1,000-2,000 bp long DNA fragments. After sonication, lysates were cleared by

microcentrifugation at 14,000 rpm. Lysates were normalized according to their OD260

measurements. Each lysate was diluted six times with dilution buffer (F-buffer, 0.5 mM

PMSF, 100 ng/ml leupeptin and aprotinin) and normalized by dilution in IP buffer

(0.015% SDS, F-buffer 0.5 mM PMSF, and 100 ng/ml leupeptin and aprotinin). The

normalized chromatin lysates were pre-cleared overnight with pre-blocked protein A/G

beads. Protein A/G beads were pre-blocked by incubation in IP buffer containing 1

µg/ml salmon sperm DNA and 1µg/ml BSA for three hours at 4°C. Precleared lysates

were incubated with 1.5 µg or 5 µL of anti-Myc (N262; Santa Cruz) or anti-Gal4 (sc-429;

Santa Cruz) respectively, at 4°C overnight. Beads were washed and eluted, followed by

reversal of the cross-linking (Kuo and Allis 1999) and DNA isolation using the Qiagen

"PCR purification kit". Eluted material was subjected to quantitative PCR amplification.

PCR was performed using the following sets of primers: human mSHMT promoter

specific primers: SC-25 (CGAGTTGCGATGCT-GTACTTCTCT) and SC-26

(GCTCGGTTGCATCATCTGCA); human cSHMT promoter specific primers: cprSHT5

(AGCGACAGGCTTAAGTGAGG) and cprSHT3 (GTGCCACCAGTCCCAGAC);

104

human TERT-specific primers: SC-9 (AGTGGATTCGCGGGCACAGA) and SC-10

(AGCACCTCGCGGTAGTGGCT); human chromosome 22-specific primers:

Chr22a (TTACAGGTAAGCACCTCCATGACC) and Chr22b:

(GCAAAAGCTACCATTTAGGAACCC); rat mSHMT-specific primers RmSHMTs

(CTGGATGACCAGTGGAAAGG) and RmSHMT (GACCCTTGCGTGATGAAAGT);

rat nucleolin-specific primers: NUCLa (CTGGGAGGGCGATGTAGAGT) and NUCLb

(GGAAGGGGGTTATCTCGAAG); rat glucokinase- specific primers: GLUa

(TGCCCGATTTTCATCTTCTT) and GLUb (CCAAGGACTTCCGCACTAAC);

To normalize samples by the amount of nonspecific DNA, we amplified a region in the

promoter of rat PCNA gene: PCNa (CGAAGCACCCAGGTAAGTGT) and

PCNb (ATCGTATCCGTGGTTTGAGC), or in the third intron of the human β-globin

gene: SC-46 (ATCTTCC-TCCCACAGCTCCT) and SC-47

(TTTGCAGCCTCACCTTCTTT). One of the primers in each pair was end-labeled with

γ32P-ATP. Amplification was performed in a T3 Thermocycler (Biometra®) for 31 cycles

at 940C for 45 sec, 600C for 45 sec, and 720C for 60 sec, followed by a final extension step

at 720C for 5 min. The optimal number of cycles for exponential amplification was

determined by kinetic analysis. PCR products were resolved on a 4% or 6%

polyacrylamide gel. PCR products were quantitated and normalized using the Molecular

Dynamics Phoshphor Imaging System.

Acknowledgements We thank Marcelo Wood for RNA samples, Julie Goodliffe for the critical reading of the manuscript and Patrick Stover for helpful discussions. This work was supported by grants from the NIH to MDC and JMS.

105

CHAPTER 4: TRRAP-dependent and TRRAP-independent

Transcriptional Activation by Myc Family Oncoproteins

The work presented in the following chapter is published. The reference for this paper is the following and is available as a supplemental file: Nikiforov MA*, Chandriani S*, Park J*, Kotenko I, Matheos D, Johnsson A, McMahon SB, Cole MD 2002. TRRAP-dependent and TRRAP-independent transcriptional activation by Myc family proteins. Molecular and Cellular Biology 22: 5054-5063. *Authors equally contributed.

106

Results Considerable effort has been expended to define a collection of genes that mediate Myc

function but few studies have addressed the mechanism by which Myc activates individual

target genes. Mutations that disrupt the evolutionarily conserved MBII region in the

transactivation domain of c-Myc have generally been found to have unaltered

transactivation in transient assays (Bello-Fernandez et al. 1993) (Kuttler et al. 2001), yet

these same mutants are defective in cell-transformation assays (Brough et al. 1995) (Stone

et al. 1987) . On the other hand, the L-Myc protein, which possesses an intact MBII

domain, is also severely defective for oncogenesis in most assays (Birrer et al. 1988). Gene

knockouts of L-myc are viable, unlike the knockouts of c-myc and N-myc (Davis et al.

1993) (Hatton et al. 1996) (Sawai et al. 1993) (Stanton et al. 1992). We were interested in

determining if chromatin modifying factors interacting with the Myc family proteins were

required for Myc-dependent activation of specific cellular promoters in the context of their

native chromosomal location in normally growing cells, and if different Myc family

proteins exhibited differential gene activation. We focused on several promoters (TERT,

CAD, HSP60, and CDK4) that have binding sites for Myc/Max heterodimers and that can

be directly crosslinked to Myc in vivo.

Myc Box II is required for activation of the TERT gene in IMR90 cells.

As an initial assay for the activation of Myc regulated genes, we tested gene expression in

response to wt c-Myc and the c-Myc(∆129-145) mutant (c-MycMBII∆ hereafter) that is

defective for oncogenic activity. The TERT gene is silent in primary human lung

fibroblasts (IMR-90) cells, and ectopic expression of wt c-Myc activates expression as

described previously (Wang et al. 1998). In contrast, the c-MycMBII∆ mutant was

107

completely defective in TERT activation (Figure 4.1A). Unlike TERT, expression of the

CAD, CDK4, and HSP60 genes is readily detectable in IMR-90 cells, and expression is

enhanced approximately 3-fold by ectopic wt c-Myc. Interestingly, the c-MycMBII∆

protein also induced CAD, CDK4, and HSP60 expression, although to a lesser extent

(Figure 4.1A). These data demonstrate that the ability of c-Myc to activate a silent gene

like TERT is completely dependent on MBII, whereas other genes exhibit only a modest

MBII-dependence.

We next examined expression of all four genes in an established fibroblast cell line

in which the c-myc-genes were knocked out by homologous recombination (Mateyak et al.

1997). Unlike in IMR-90 cells, the TERT gene is expressed even in the absence of

endogenous c-Myc in c-myc-null fibroblasts (Figure 4.1A). TERT is induced by

reconstitution with either wt c-Myc (5.4 fold) or c-Myc∆ (3 fold). The response of the

CAD, CDK4, and HSP60 genes was similar to that in IMR-90 cells, with approximately 3-

fold and 2-fold induction by wildtype c-Myc and by c-MycMBII∆, respectively. Thus,

TERT, CAD, CDK4 and HSP60 expression is sustained in immortalized cell lines in the

absence of c-Myc, and the modest Myc-dependent activation of these genes does not

depend on the MBII domain.

L-Myc is a weak activator of TERT, but not of other Myc targets in IMR90 cells.

We were next interested in whether other Myc family proteins had a similar capacity to

activate endogenous promoters. L-Myc has been described as having both weak oncogenic

activity and poor activation in transient assays (Barrett et al. 1994) (Birrer et al. 1988). We

transduced a FLAG-tagged L-Myc protein into IMR90 cells, which was expressed at the

same level as c-Myc by Western blotting (Figure 4.1B). L-Myc reproducibly activated the

108

TERT gene to approximately 20% of the level induced by c-Myc (Figure 4.1B). On the

other hand, other Myc targets (CAD, CDK4, and HSP60) were up-regulated by L-Myc to

the same extent as by c-Myc. In contrast to its response in transient assays, the activity of

L-Myc in reconstituted c-myc-null fibroblasts paralleled that of c-Myc where all four Myc

target genes were induced 2-3 fold. Thus, L-Myc exhibits similar activation activity as c-

Myc, with the noteworthy exception of activation of the TERT gene in primary human

fibroblasts.

TERT activation correlates with TRRAP binding.

The data above establish a strong requirement for the MBII of c-Myc for TERT activation

in primary cells. Previously we reported that the MBII domain in c- and N-Myc plays an

important role in the interaction of Myc with several nuclear proteins including TRRAP,

TIP48/49 and BAF53 proteins (McMahon et al. 1998) (Park et al. 2002) (Wood et al. 2000).

We were interested in establishing a direct link between recruitment of individual cofactors

by the Myc family proteins and the activation of Myc-responsive genes. We first tested if

L-Myc bound to the same cofactors as other Myc family proteins. To this end, FLAG-

tagged L-Myc and N-Myc proteins were transiently expressed in HEK293 cells and tested

for binding to endogenous nuclear cofactors. L-Myc binding to TRRAP and TIP49 were

substantially reduced compared to that of N-Myc (Figure 4.2). However, binding of

nuclear cofactors to L-Myc was still detectable in comparison to N-MycMBII∆, which

showed no binding to the nuclear cofactors tested. This weak binding of L-Myc to TRRAP

correlates well with its impaired ability to activate the silent TERT gene in IMR90 cells,

but, as with c-MycMBII∆, it does not correlate with the ability of L-Myc to activate other

Myc-target genes.

109

BA

CAD

CD

K 4

HSP

60

GA

PDH

TER

T

α-F

LAG

RT-

PCR

myc

+/+

prim

ary

hum

anfib

robl

asts

myc

-/-im

mor

taliz

ed ra

tfib

robl

asts

VC

L

12.

72.

5

12.

41.

814.

3

12.

32.

9

VC

L

12.

63.

5

3.0

14.

5

12.

83.

2

12.

13.

1

VC

C∆ 3.0

15.

4

12.

23.

5

12.

03.

3

12.

31.

8

VC

C∆

12.

92.

0

12.

82.

1

12.

71.

5

myc

+/+

prim

ary

hum

anfib

robl

asts

myc

-/-im

mor

taliz

ed ra

tfib

robl

asts

BA

CAD

CD

K 4

HSP

60

GA

PDH

TER

T

α-F

LAG

RT-

PCR

myc

+/+

prim

ary

hum

anfib

robl

asts

myc

-/-im

mor

taliz

ed ra

tfib

robl

asts

VC

L

12.

72.

51

2.7

2.5

12.

41.

81

2.4

1.81

4.3

4.3

12.

32.

91

2.3

2.9

VC

LV

CL

12.

63.

51

2.6

3.5

3.0

14.

53.

01

4.5

14.

5

12.

83.

21

2.8

3.2

12.

13.

11

2.1

3.1

VC

C∆ 3.0

15.

4

12.

23.

5

12.

03.

3

12.

31.

8

VC

C∆

12.

92.

0

12.

82.

1

12.

71.

5

myc

+/+

prim

ary

hum

anfib

robl

asts

myc

-/-im

mor

taliz

ed ra

tfib

robl

asts

VC

C∆ 3.0

15.

4

12.

23.

5

12.

03.

3

12.

31.

8

VC

C∆ 3.0

15.

41

5.4

12.

23.

5

12.

03.

3

12.

31.

8

VC

C∆

12.

92.

0

12.

82.

1

12.

71.

5

VC

C∆

12.

92.

01

2.9

2.0

12.

82.

11

2.8

2.1

12.

71.

51

2.7

1.5

myc

+/+

prim

ary

hum

anfib

robl

asts

myc

-/-im

mor

taliz

ed ra

tfib

robl

asts

Figu

re 4

.1 A

ctiv

atio

n of

end

ogen

ous g

enes

by

c-M

yc, c

-Myc∆

and

L-m

yc

110

Figure 4.1 Activation of endogenous genes by c-Myc, c-Myc∆ and L-mycIMR 90 and HO15.19 cells were infected with empty vector or retroviral vectors expressing FLAG-tagged c-myc or c-myc∆ cDNAs (A) or empty vector and c-myc or L-myc cDNA (B). To determine protein expression, normalized cell lysates were immunoprecipitated with anti-FLAG antibodies, resolved on 8% SDS polyacrylamidegel followed by Western blotting with anti-FLAG specific antibodies. Quantitative PCR was performed with cDNA synthesized on total RNA isolated from infected cells using radiolabelled oligos specific for human or rat genes indicated in the center. Quantitation of PCR products was performed using Molecular Dynamics PhosphorImaging System. Numbers below a panel indicate the fold induction determined by dividing a TERT-, CDK4-, HSP60- or CAD- specific signal by a corresponding GAPDH-specific signal and by the ratio of these signals in the "vector" lane.

111

220

66

46

V L N∆N

α-TRRAP

α-FLAG

α-Tip 49

220

66

46

V L N∆N

220

66

46

V L N∆N

α-TRRAP

α-FLAG

α-Tip 49

Figure 4.2 Recruitment of nuclear cofactors by L-Myc and N-Myc.HEK 293 cells were transiently transfected with FLAG-tagged constructs encoding proteins shown on the top. Normalized cell lysates were prepared and immunoprecipitated with anti-FLAG antibodies. Immunoprecipitated complexes were resolved on a gradient 4%-15% SDS polyacrylamide gel and probed by western blotting with several antibodies, indicated on the left. Protein molecular weights are shown on the right in kDa.

112

E1A-N-Myc∆ chimeras selectively bind to TRRAP.

Further support for a link between TRRAP and TERT activation came from an unexpected

direction through studies of the adenoviral oncoprotein E1A. E1A shares many of the

biological activities of c-Myc, including oncogenic transformation of primary cells in

cooperation with the H-rasG12V oncogene, blocking differentiation, and promoting

apoptosis (Flint and Shenk 1997). E1A is thought to predominantly act through the E2F

family of transcription factors (Reichel et al. 1987). We previously showed that TRRAP

binds to the E2F transactivation domain (McMahon et al. 1998), and TRRAP recruitment is

important for E2F activity (Lang et al. 2001). Since E1A and E2F bind to many of the

same cellular cofactors, we tested if E1A binds to TRRAP. As shown in Figure 4.3A,

immunoprecipitates of E1A from HEK293 cells indeed contained TRRAP protein when

assayed by Western blotting. The biological consequences of E1A binding to TRRAP are

presently unknown, but E1A may modify SAGA function in adenovirus infected cells or

bind to other TRRAP-containing complexes (Fuchs et al. 2001).

It was previously shown that chimeric proteins with the E1A N-terminus fused to the

c-Myc DNA binding domain were potent oncogenes (Ralston 1991) . The binding of E1A

to TRRAP prompted us to test if this interaction could explain the oncogenic activity of

E1A-Myc chimeras and be used to complement the defect in TRRAP binding of the Myc

Box II mutations in Myc. We used a FLAG-tagged N-Myc expression vector since N-Myc

binds avidly to TRRAP, and we also derived a deletion mutant in the N-Myc MBII-domain

that is defective in both TRRAP binding and oncogenic activity (Figure 4.2 and see below).

We then constructed a series of chimeras in which various segments of the E1A N-terminus

were fused to the N-MycMBII∆ protein (Figure 4.3B).

113

N-Myc∆

N-MycMBII DNA binding

4391

1

10-39 / 1

10-33 / 1

E1A 10-39 / N-Myc∆

E1A 10-33 / N-Myc∆

V

V

V

B

Alys

atecontro

l

α-E1A

α-TRRAP220

C

α-TRRAP

α-N-myc

α-Tip 49

V N 10-3910-33

N∆

220

66

46

N-Myc∆


4391

1

10-39 / 1

10-33 / 1

E1A 10-39 / N-Myc∆

E1A 10-33 / N-Myc∆

V

V

V

N-Myc∆


4391

1

10-39 / 1

10-33 / 1

E1A 10-39 / N-Myc∆

E1A 10-33 / N-Myc∆

V

V

V

B

Alys

atecontro

l

α-E1A

α-TRRAP220

lysate

control

α-E1A

α-TRRAP220

C

α-TRRAP

α-N-myc

α-Tip 49

V N 10-3910-33

N∆

220

66

46

Figure 4.3 E1A-N-Myc∆ chimeras selectively precipitate TRRAP from 293 cells

114

Figure 4.3 E1A-N-Myc∆ chimeras selectively precipitate TRRAP from 293 cellsA. Normalized whole cell lysates were prepared from HEK 293 cells which constitutively express the adenovirus E1A protein. Immunoprecipitates were resolved on SDS polyacrylamide gel and probed for the presence of the TRRAP protein in western blotting using home made rabbit anti-TRRAP antisera. Results of these studies demonstrate that the TRRAP protein specifically associates with the E1A protein in vivo. B. Schematic representation of the E1A-N-Myc∆ fusion proteins. E1A portion of the chimera, Myc Box II (MBII) domain, and DNA-binding domain are shown by meshed, hatched and dotted boxes, respectively. C. Recruitment of nuclear cofactors by N-Myc-derived proteins. HEK 293 cells were transiently transfected with FLAG-tagged constructs encoding proteins shown in (B). Normalized cell lysates were prepared and immunoprecipitated with anti-FLAG antibodies. Immunoprecipitated complexes were resolved on a gradient 4%-15% SDS polyacrylamide gel and probed by Western blotting with several antibodies, indicated on the left. Protein molecular weights are shown on the right in kDa.

115

We first tested if a specific E1A domain could restore TRRAP binding to N-

MycMBII∆. Proteins were transiently expressed in HEK293 cells, immunoprecipitated

through a FLAG epitope tag, and assayed for TRRAP binding by Western blotting.

Preliminary mapping demonstrated that E1A sequences 1-86 could restore TRRAP binding

(data not shown), which was subsequently narrowed to E1A sequences 10-39 (Figure 4.3C),

although binding was consistently weaker than to wildtype N-Myc. Fusion of E1A amino

acids 10-33 to N-MycMBII∆ failed to restore binding (Figure 4.3C), implicating E1A

amino acids 33-39 as critical for TRRAP interaction. We also tested the E1A-N-Myc

chimeras for binding to another Myc cofactors, TIP49. There was no enhancement of

TIP49 binding to the E1A-N-Myc∆ chimeras compared to N-MycMBII∆ (Figure 4.3C).

Thus, a small E1A domain selectively reconstitutes binding of TRRAP to an N-Myc

mutant with a deletion of the Myc Box II region, but not binding to another Myc-associated

complex(es).

E1A-Myc∆ chimeras activate the silent TERT gene.

The E1A-N-Myc∆ chimeras, wt N-Myc, and N-Myc∆ were each introduced into

IMR-90 and c-myc null fibroblasts to assay their ability to activate target gene expression.

Remarkably, the E1A(10-39)-N-MycMBII∆ fusion protein that bound to TRRAP could

also activate the silent TERT gene in IMR90, similar to wildtype N-Myc (Figure 4.4A). In

contrast, the N-MycMBII∆ mutant and the E1A(10-33)-N-MycMBII∆ fusion proteins were

completely defective in TERT activation. The response of the endogenous CAD gene to

N-Myc and the E1A-N-Myc chimeras was similar to that for c-Myc in IMR90 cells. A

representative experiment (Figure 4.4A) demonstrated that N-Myc provided a 3.2 fold

activation of CAD, whereas N-MycMBII∆ induced CAD 2.1 fold. Each of the E1A-N-

116

Figu

re 4

.4 E

1A-N

-Myc∆

chim

eras

res

tore

the

tran

sact

ivat

ion

pote

ntia

l of N

-Myc∆

A

RT-

PCR

α-N

-Myc

myc

+/+

prim

ary

hum

an fi

brob

last

sm

yc -/-

imm

orta

lized

rat f

ibro

blas

ts

TER

T

CA

D

GA

PDH

B

1.0

3.2

2.1

3.0

1.9

VN

10-39

10-33

N∆

1.0

4.7

3.6

5.2

4.3

1.0

4.5

2.2

3.9

3.0

VN

10-39

10-33

N∆

A

RT-

PCR

α-N

-Myc

myc

+/+

prim

ary

hum

an fi

brob

last

sm

yc -/-

imm

orta

lized

rat f

ibro

blas

ts

TER

T

CA

D

GA

PDH

B

1.0

3.2

2.1

3.0

1.9

1.0

3.2

2.1

3.0

1.9

VN

10-39

10-33

N∆

VN

10-39

10-33

N∆

1.0

4.7

3.6

5.2

4.3

1.0

4.7

3.6

5.2

4.3

1.0

4.5

2.2

3.9

3.0

1.0

4.5

2.2

3.9

3.0

VN

10-39

10-33

N∆

VN

10-39

10-33

N∆

117

Figure 4.4 E1A-N-Myc∆ chimeras restore the transactivation potential of N-Myc∆IMR 90 (A) and HO15.9 (B) cells were infected with the retroviral vectors carrying cDNAs for proteins shown in Figure 4.3B. To determine protein expression, normalized lysates from infected cells were immunoprecipitated with anti-FLAG antibodies, resolved on 8% SDS polyacrylamide gel followed by Western blotting with N-Myc-specific antibodies. Quantitative PCR was performed on cDNA synthesized on RNA isolated from cells expressing the constructs indicated on the top. PCR was done using radiolabelled oligos specific for cDNA of human (A) or rat (B) genes shown in the center. Quantitation of PCR was done using Molecular Dynamics Phosphor Imaging System. Numbers below a panel indicate a fold induction determined by dividing a TERT- or CAD- specific signal by a corresponding GAPDH-specific signal and by the ratio of these signals in the "vector" lane. A representative experiment is shown.

118

MycMBII∆ chimeras activated CAD to a similar extent, including the E1A(10-33)-N-

MycMBII∆ fusion that fails to activate TERT in IMR90. All proteins were expressed at

similar levels, although N-Myc∆ was reproducibly expressed at a higher level than

wildtype N-Myc and the E1A chimeras were expressed at a lower level (Figure 4.4A, top

panel). These data provide a further link between recruitment of TRRAP by Myc and its

ability to activate transcription of the silent TERT locus.

In c-myc-null fibroblasts, the response of the CAD and TERT genes to the N-Myc

and E1A-N-MycMBII∆ proteins was slightly higher than that in primary human fibroblasts.

As shown in Figure 4.4B, expression of CAD and TERT was induced 4.5 and 4.7 fold by

N-Myc, and 2.2 and 3.6 fold, respectively, by N-MycMBII∆. The E1A(10-39)- and

E1A(10-33)-N-MycMBII∆ proteins enhanced CAD and TERT expression to a level similar

to that induced by wildtype N-Myc. There was no correlation between TERT or CAD

activation in c-myc-null fibroblasts by fusion proteins that can activate TERT, E1A(10-39)-

N-MycMBII∆, and cannot activate TERT, E1A(10-33)-N-MycMBII∆, in IMR90 cells

(compare Figures 4.4A and 4.4B).

It is noteworthy that the ability of N-MycMBII∆ to induce transcription of basally

expressed genes in c-myc-null fibroblasts (exemplified by the CAD and TERT genes)

correlates well with its ability to complement the growth defect of these cells. As shown in

Figure 4.5A, N-Myc completely rescues the slow growth phenotype of c-myc-null cells,

decreasing their doubling time from 45 HRS to 16 HRS. Expression of N-MycMBII∆ also

resulted in a decrease of the doubling time to approximately 26 HRS. The E1A-N-

MycMBII∆ chimeras both restored the growth rate of c-myc-null cells to levels that are

intermediate between N-Myc and N-MycMBII∆. Expression of E1A(10-39)-N-Myc∆

119

02468

101214161820

Vector C N∆ 10-39 10-33

E1A-N∆ fusions

L N

Foci

/dis

h

B

0

10

20

30

40

50

60

Vector C N∆ 10-39 10-33

E1A-N∆ fusions

L N

Dou

blin

g tim

e (h

r)

A

02468

101214161820

Vector C N∆ 10-39 10-33

E1A-N∆ fusions

L N

Foci

/dis

h

B

0

10

20

30

40

50

60

Vector C N∆ 10-39 10-33

E1A-N∆ fusions

L N

Dou

blin

g tim

e (h

r)

A

Figure 4.5 Distinct phenotypes of c-Myc, L-Myc, N-Myc and E1A-N-Myc∆chimeras in cell growth and primary cell transformationA. Doubling times of myc-null cells reconstituted with the indicated Myc-derived proteins. B. Oncogenic activity of the Myc family proteins and E1A-N-Myc∆ chimeras in primary rat embryo cells. c-myc, L-myc, N-myc, N-myc∆, or E1A-N-myc∆ chimeras were cotransfected with H-rasG12V into early passage rat embryo fibroblasts and foci were scored after 18 days.

120

provided a doubling time of 21+0.3 hr versus 26±5 HRS for E1A(10-33)-N-Myc∆.

Interestingly, L-Myc, despite poor binding to TRRAP, restored the growth rate of c-myc-

null cells to a level nearly identical (20 HRS) to that resulted from the expression of c-Myc

or N-Myc (Figure 4.5A).

N-Myc, c-Myc and E1A-N-MycMBII∆ proteins recruit HAT activity to the

TERT promoter in IMR90 cells

The data above strongly suggest that recruitment of a TRRAP-containing

complex(es) by Myc is required for activation of the TERT gene in primary human

fibroblasts. Several recent studies have demonstrated that TRRAP may exist in a

complex with GCN5, a histone H3-specific acetyltransferase (McMahon et al. 2000)

(Park et al. 2001), TIP60, a histone H4-specific HAT(Ikura et al. 2000), or a novel

undefined H4 HAT complex (Park et al. 2002). Moreover, recruitment of H4 (but not H3)

HAT activity correlated with the activation of several Myc-responsive genes in starved

cells after serum stimulation (Frank et al. 2001). We were interested in whether Myc is

able to recruit a specific HAT activity to the nucleosomes on the TERT promoter in

IMR90 cells. We first tested if interactions with TRRAP are crucial for recruiting HAT

activity by Myc. To this end, complexes containing FLAG-tagged N-Myc, N-

MycMBII∆, and E1A(10-39)-N-MycMBII∆ proteins were transiently expressed in HEK

293 cells, immunoprecipitated using FLAG-specific antibodies, eluted from the beads,

and assayed for HAT activity and the presence of TRRAP. As shown in Figure 4.6A, N-

Myc co-immunoprecipitates TRRAP and both H3- and H4-specific acetyltransferase

activity. In contrast, the N-Myc∆ protein fails to recruit TRRAP and any HAT activity,

whereas E1A(10-39)-N-Myc∆ recruits both TRRAP (Figure 4.2C) and H3-/H4-HATs.

121

The HAT activity recruited by E1A(10-39)-N-Myc∆ was not as robust as for wt N-Myc,

consistent with the lower amount of TRRAP recruited by this fusion. L-Myc also

recruited a reduced level of HAT activity compared to N-Myc, but the level was higher

than for N-MycMBII∆ (Figure 4.6A).

We were next interested in whether N-Myc and E1A(10-39)-N-MycMBII∆

recruited histone H3- and/or H4-specfic acetyltransferase activity to the TERT promoter.

Antibodies specific to acetylated histones H3 or H4 were used to immunoprecipitate

cross-linked chromatin from IMR90 cells expressing N-Myc, N-MycMBII∆, E1A(10-

39)-N-MycMBII∆, c-Myc, c-MycMBII∆ or vector. After reversal of the cross-linking,

DNA was purified and subjected to quantitative PCR with oligos corresponding to the

region of the TERT promoter encompassing two Myc-binding sites, or to a region of

similar size in the third intron of a silent ß-globin gene for a control purposes. As shown

in Figure 4.6B, there is a significant enrichment of the TERT promoter sequences (3.9-

6.1 fold) in the samples immunoprecipitated with anti-acetylated-H3 and anti-acetylated-

H4 antibodies from all cells in which the TERT gene was activated (c-Myc, N-Myc or

E1A10-39-N-MycMBII∆) compared to empty vector, using ß-globin as a normalization

control. There was very little or no enrichment in samples obtained from cells expressing

c-Myc∆ or N-MycMBII∆, in which the TERT gene is silent. We assume that the signal

from the ß-globin gene represents a measure of nonspecific DNA binding to the beads

during the immunoprecipitation step, since this gene should be silent in primary

fibroblasts. However, we cannot exclude the possibility that there is a low level of

histone acetylation even on silent genes, including the silent TERT locus. Taken together,

our data demonstrate that Myc-dependent activation of the silent TERT gene in IMR90

122

A

B

α-FLAG 66

HATActivity

V N 10-39N∆ V L N∆N

H3H4

α-N-Myc

TERT

GAPDHRT-PCR

1.0 6.1 1.4 3.9

TERT

1.0 4.3 1.8

TERT

ß-globin

ß-globin

V N 10-39N∆V C C∆

1.0 5.6 1.4 1.0 5.5 1.0 4.7

α-c-Myc

α-Acetyl-H3ChIP

α-Acetyl-H4ChIP

A

B

α-FLAG 66

HATActivity

V N 10-39N∆V N 10-39N∆ V L N∆NV L N∆N

H3H4

α-N-Myc

TERT

GAPDHRT-PCR

1.0 6.1 1.4 3.9

TERT

1.0 4.3 1.8

TERT

ß-globin

ß-globin

V N 10-39N∆V C C∆

1.0 5.6 1.4 1.0 5.5 1.0 4.7

α-c-Myc α-N-Myc

TERT

GAPDHRT-PCR

1.0 6.1 1.41.0 6.1 1.4 3.9

TERT

1.0 4.3 1.8

TERT

ß-globin

ß-globin

V N 10-39N∆V C C∆

1.0 5.6 1.4 1.0 5.5 1.0 4.71.0 5.6 1.4 1.0 5.5 1.0 4.7

α-c-Myc

α-Acetyl-H3ChIP

α-Acetyl-H3ChIP

α-Acetyl-H4ChIP

α-Acetyl-H4ChIP

Figure 4.6 Recruitment of histone acetyltransferase activity by Myc family proteins

123

Figure 4.6 Recruitment of histone acetyltransferase activity by Myc family proteins(A) Complexes containing the transiently expressed FLAG-tagged proteins designated on the top were immunoprecipitated from lysed HEK 293 cells using anti-FLAG antibodies. FLAG-tagged proteins were eluted from the beads and assayed for HAT activity on equal amounts of core histones isolated from HeLa cells. Reaction mixtures were resolved on gradient 4%-15% SDS polyacrylamide gels and transferred to a nitrocellulose membrane. The upper section of the membrane was probed by western blotting with the antibodies designated on the left. Protein molecular weights are shown on the right. The bottom portion of the membrane was subjected to fluorography for HAT activity. Positions of histones H3 and H4 are shown on the right. B. IMR-90 cells were infected with retroviral vectors carrying the cDNAs indicated on the top. To determine protein expression, normalized lysates from infected cells were immunoprecipitated with anti-FLAG antibodies, resolved on 8% SDS polyacrylamidegel, followed by Western blotting with anti-N-Myc antibodies. cDNA was synthesized on RNA isolated from cells infected with the constructs indicated above, and used as a template in PCR with radiolabelled oligos specific for human TERT or GAPDH cDNA. Cells infected with the constructs indicated above were cross-linked and lysed, and chromatin was immunoprecipitated with either anti-acetylated H3- or anti-acetylated H4 histone antibodies followed by the reversion of the cross-linking and DNA isolation. Isolated DNA was used in PCR with radiolabelled oligos flanking Myc binding sites in the promoter of human TERT gene or a DNA region of similar size in the third intron of a silent �-globin gene. Quantitation of PCR products was performed using Molecular Dynamics Phosphor Imaging System. Numbers below a panel indicate a fold difference in a signal intensity determined by dividing a TERT- specific signal by a corresponding globin-specific signal and by the ratio of these signals in the "vector" lane.

124

cells requires recruitment of TRRAP containing complexes to the TERT promoter and

that this recruitment is accompanied by increased acetylation of histones H3 and H4 in

the promoter area.

TRRAP binding correlates strongly with Myc family oncogenic activity.

One of the most revealing assays for Myc oncogenic function is the transformation of

primary rodent cells in cooperation with the H-rasG12V oncogene (Land et al. 1983). This

assay was originally used to demonstrate the potent transforming activity of E1A-Myc

chimeras (Ralston 1991) , so we were interested in determining if the E1A-MycMBII∆

proteins described above would also transform cells and how this activity related to that of

L-Myc. Each expression construct was cotransfected with H-rasG12V into early passage

rat embryo fibroblasts (Land et al. 1983), and foci of transformed cells were scored after

18-21 days. As shown in Figure 4.5B, c-Myc and N-Myc are highly oncogenic, whereas

deletion of the N-Myc MBII abolishes transforming activity. L-Myc has a significantly

weaker transforming activity, consistent with earlier reports (Birrer et al. 1988). Fusion of

E1A sequences 10-39 to N-MycMBII∆ partially rescued N-Myc transforming activity. In

contrast, fusion of E1A(10-33) to N-MycMBII∆ was as defective as N-MycMBII∆ (Figure

4.5B). Thus, mapping of E1A segments that rescue the oncogenic activity of Myc in

primary cells identifies the same region (10-39) as that required for activation of the silent

TERT gene, and the defective E1A(10-33) fusions define amino acids 33-39 as a critical

boundary for both activities.

The fusion of E1A sequences to N-MycMBII∆ mutants restored binding to TRRAP,

but we wanted to demonstrate a more direct involvement of TRRAP in the biological

activity of the chimeras. We previously described a segment of TRRAP (amino acids

125

1261-1579) that is a dominant inhibitor of Myc oncogenic activity when ectopically

expressed in conjunction with the H-rasG12V oncogene. Cotransfection of plasmids

expressing TRRAP(1261-1579) and E1A(10-39)-N-MycMBII∆ proteins strongly inhibited

focus formation in REFs (90%) compared to cotransfection of E1A(10-39)-N-MycMBII∆

with empty vector (data not shown). These results support the co-immunoprecipitation

experiments and argue for a direct involvement of TRRAP in the oncogenic activity of

E1A-MycMBII∆ chimeras.

126

Discussion

Recruitment of TRRAP by Myc family proteins.

It is has long been known that Myc-mediated cellular phenotypes largely depend on

the integrity of the Myc transactivation domain, particularly Myc Box II. Several recent

studies defined nuclear proteins capable of interaction with MBII, thus providing molecular

insight into mechanisms of Myc action (McMahon et al. 1998) (Wood et al. 2000).

TRRAP, an integral component of human SAGA and TIP60 complexes (Ikura et al. 2000);

(Vassilev et al. 1998), was the first polypeptide shown to interact with MBII (McMahon et

al. 1998). Human SAGA-related complexes contain GCN5/PCAF and specifically

acetylate histone H3, whereas the TIP60 complex acetylates predominantly histone H4.

Recently, we found another TRRAP-containing complex with H4 HAT activity that does

not contain TIP60 (Park et al. 2002). Results obtained from histone acetylation in vitro and

chromatin immunoprecipitation in cultured cells strongly suggest that Myc interacts with at

least two distinct TRRAP containing complexes containing HATs with different histone

specificities. A third TRRAP-containing complex that lacks HAT activity has also been

described recently (Fuchs et al. 2001). Other Myc interacting proteins include TIP48,

TIP49 and BAF53 (Park et al. 2002) (Wood et al. 2000) may have a functional role in other

TRRAP-containing (Ikura et al. 2000) or non-TRRAP complexes (G. LeRoy and G. Wang,

personal communications). In our study, it was important to determine the individual

involvement of these proteins in Myc-dependent pathways. We found that L-Myc interacts

with TRRAP/TIP49 polypeptides and recruits HAT activity much more weakly than c- or

N-Myc, which correlates well with the weak oncogenic activity of L-Myc in the H-ras-

cooperation assay.

127

We used a small fragment of adenoviral E1A protein (amino acids 10-39) to

selectively restore an interaction with TRRAP in the context of a MBII deletion without

detectable concomitant binding to TIP49 (Figure 4.3C). This portion of E1A does not

overlap the conserved CR1, CR2, and CR3 regions implicated in other protein-protein

interactions and E1A functions (Barbeau et al. 1992). A previous study of E1A-interacting

proteins identified a 400 kDa band in immunoprecipitates from adenovirus infected cells

that was dependent on amino acids 4-49 (Barbeau et al. 1994), overlapping the 33-39

boundary that defines TRRAP recruitment to the E1A/N-MycMBII∆ fusion proteins. An

E1A mutant with a deletion of amino acids 26-35 was defective for p400 binding but

retained the ability to transform baby rat kidney cells in combination with H-rasG12V

(Barbeau et al. 1994). However, more detailed analysis of the E1A∆26-35 mutant reveals a

substantial deficiency in transforming activity and TRRAP binding compared to wt E1A

(Deleu et al. 2001). It is not yet clear if p400 is TRRAP or a recently described Swi/Snf-

related protein of similar migration (Fuchs et al. 2001). It is tempting to speculate that E1A

binding to TRRAP might alter its activity or the activity of TRRAP-containing chromatin

modifying complexes. E1A itself does not activate TERT expression (Wang et al. 1998),

presumably because it lacks a DNA binding domain that would target the protein to the

TERT locus.

An array of complexes have been found to contain TRRAP and/or bind to Myc

proteins (reviewed in the Introduction). At face value, the observation that the E1A(10-

39)-N-MycMBII∆ protein binds to TRRAP but not TIP49 argues that the chimera does not

bind to either the TIP60 or p400 complexes, both of which contain TIP49 (Fuchs et al.

2001) (Ikura et al. 2000). We recently showed that a TRRAP-associated H4 HAT complex

128

can be distinguished biochemically from the TIP60 complex, although the enzyme itself

has not been identified (Park et al. 2002). It is tempting to speculate that this latter

complex is recruited by the E1A-N-MycMBII∆ chimera.

TRRAP-dependent activation of Myc-responsive genes

Telomerase is a two-component ribonucleoprotein DNA polymerase composed of

telomerase RNA (RT) and telomerase reverse transcriptase (TERT) (reviewed in (Greider

1999). Telomerase synthesizes short DNA repeats at chromosomal ends and is required for

maintaining genomic stability (reviewed in (Artandi and DePinho 2000). Although

telomerase is active in many fetal tissues during embryonic development, its activity is

gradually repressed in most somatic cells shortly after birth. However, the TERT gene

becomes active again during cell immortalization and in the vast majority of human cancers

(Greider 1999). Despite an extensive search for TERT transcriptional activators, Myc is

the only protein that has been shown to activate expression of a silent TERT gene by

binding to its promoter (Wang et al. 1998). However, the molecular mechanism of this

activation by Myc was not elucidated. We showed that overexpression of c-Myc, N-Myc,

and to a much lesser extent L-Myc activates expression of the silent TERT gene. The use

of Myc mutants and E1A-Myc fusion proteins provides compelling evidence that TERT

activation and nucleosome modification directly result from the recruitment of the TRRAP

protein and its associated histone H3 and H4 HAT activities by the Myc transactivation

domain.

An important finding from these studies is the striking difference between primary

and immortalized exponentially growing cells in their requirement for TRRAP-containing

complexes for activation of the TERT gene. In primary cells, the TERT gene is strongly

129

repressed through unknown mechanisms, whereas in immortalized cells TERT is basally

expressed even in the absence of Myc. Presumably, other transcription factors such as Sp1

are responsible for basal promoter activity in c-myc-null cells, but these proteins are

excluded from activating TERT expression in primary cells even though proteins such as

Sp1 are ubiquitously expressed. Repression of TERT may involve active histone

deacetylation since tricostatin A can activate TERT expression in different cell types and

Mad/Max complex can bind to the E-box in TERT promoter (Xu et al. 2001). However,

tricostatin A might also activate other genes that directly or indirectly derepress TERT,

including Myc itself.

In contrast to the silent TERT gene, induction of basally expressed genes, CAD,

CDK4 and HSP60 in exponentially growing IMR90 cells, and CAD, CDK4, HSP60, and

TERT in exponentially growing c-myc-null cells, did not depend strictly on the MBII

domain that is required for recruitment of TRRAP and HAT activity by Myc. This

observation is consistent with a previous study of Myc target genes in which Myc binding

to the promoter of the active TERT gene in exponentially growing neuroblastoma cell line

occurred without a concomitant change in histone acetylation (Eberhardy et al. 2000).

Moreover, recent studies demonstrated the ability of Myc to activate transcription of the

CAD gene through mechanisms independent from histone acetylation (Eberhardy and

Farnham 2001). We found that Myc and the E1A-N-MycMBII∆ proteins were sometimes

more potent than MycMBII∆ in the induction of the CAD, CDK4, HSP60, and TERT

genes (in HO15.19 cells), suggesting that TRRAP complexes may play a limited role in

transcriptional activation of these genes. Furthermore, recent reports show that TRRAP and

its associated H4 (but not H3) HAT activities may have a more significant role in the

130

induction of Myc target genes during the serum stimulation of starved cells (Freytag et al.

1990), as opposed to the exponentially growing cells studied here.

The strict dependence of TERT activation on the MBII domain and TRRAP/HAT

recruitment in primary fibroblasts correlates well with the similar dependence on the MBII

domain in the oncogenic transformation of primary cells in the Myc/H-ras cooperation

assay. This link is further supported by the weak binding of L-Myc to TRRAP and its

limited recruitment of HAT activity, which correlates with its relatively weak oncogenic

activity in primary fibroblasts. An H-ras oncogene alone can transform immortalized cells

but requires a cooperating myc oncogene to transform primary cells. Since regulation of

the CAD, CDK4, and HSP60 promoters by Myc shows only a minimal dependence on

MBII, it is tempting to suggest that genes in this class may not be the rate-limiting targets

of Myc in primary cells. On the other hand, we and others showed that L-Myc and

MycMBII∆ were capable of a complete or partial growth enhancement of c-myc-null

fibroblasts ((Bush et al. 1998) (Landay et al. 2000) and this study), and this enhancement

correlated with induction of CAD, CDK4, HSP60, and TERT genes (Figure 4.3B and

Figure 4.6A). This observation supports the idea that Myc promotes cell growth and

causes cellular transformation through activation of different sets of genes. One attractive

model is that the transformation of primary cells may require the activation of strongly

repressed TERT-like genes, which can only be achieved through HAT-mediated chromatin

modifications. However, at the present time we do not know any other Myc-responsive

genes which are silent in normal fibroblasts and activated in transformed or immortalized

cells. On the other hand, TERT itself is unlikely to be the rate-limiting Myc target because

TERT overexpression is not able to substitute for Myc in the transformation of rat

131

fibroblasts (Greenberg et al. 1999) and cells derived from mice lacking telomerase RNA

can still be transformed by cooperating myc and H-ras oncogenes (Blasco et al. 1997). A

screen for other genes that are strongly repressed in primary cells and activated by Myc

overexpression may identify critical Myc target genes involved in oncogenesis.

132


Vectors

Expression plasmids were created by standard methods in a CMV-promoter driven vector

containing FLAG-epitope tag or in retroviral expression vector pLXSH and verified by

sequence analysis. Details of individual constructs are available upon request.

Cell Culture, Transfection and Retroviral Infection

Primary human fibroblasts (IMR90), Rat embryo fibroblasts, Rat c-myc-null fibroblasts

(HO15.19), Retroviral producer PhoeNX cell line and HEK 293 cells were cultured in

DMEM supplemented with 10% calf serum.

Retroviral infection of IMR90 and HO15.19 cells was performed according to (Pear et al.

1993), using PhoeNX cells. PhoeNX cells were transfected via the calcium phosphate

method. To obtain a transduced population, after completion of infection IMR90 or

HO15.19 cells were selected for resistance to hygromycin (150 µg/ml) for 7 days.

Reverse Transcription and PCR

Total RNA was isolated from infected cells using TRIzol reagent (GIBCO-BRL). Reverse

transcription of 1-3 µg of RNA was performed using SuperScriptTM First-Strand Synthesis

System (GIBCO-BRL), according to the manufacturer instructions. PCR was performed

on cDNA using the following sets of oligos: human TERT-specific oligos: HT1

(CGGAAGAGTGTCTGGAGCAA) and HT2 (GGATGAAGCGGAGTCTGGA); human

GAPDH-specific oligos: HG1 (CTCAG-ACACCATGGGGAAGGTGA) and HG2

(ATGATCTTGAGGCTGTTGTCATA); human CDK4-specific oligos: HCD1

(GTGGACATGTGGAGTGTTGG) and HCD2 (GCCCTCTCAGTGTCCAGAAG); human

133

HSP60-specific oligos: HHSP1 (ACAAGTGATGTTGAAGTGAATG) and HHSP2

(ATTGCTGGAATTTTGAGTGTTC); rat TERT-specific oligos: RT1

(CGTAAGAGTGTGTGGAGCAA) and RT2 (GGATGAAGCGCAGTCTGCA); rat

GAPDH-specific oligos: RG1 (AACGGAT-TTGGCCGTATTGGCCG) and RG2

(GACAATCTTGAGGGAGTTGT-CATA); rat CDK4-specific oligos: RCD1

(TGGGTGCGGTGCCTATGGGA) and RCD2 (CGCATCTGGTAGCTGTAGATTC); rat

HSP60-specific oligos: RHSP1 (ACAAGTGATGTTGAAGTGAATG) and RHSP2

(ATTGCAGGAATTTTAAGTGCTC). Oligos specific for human and rat CAD cDNA:

CAD1 (GGTGGATCTGGAG-CATGAGTGGA) and CAD2

(AGATGGAAGCGGCCATCAGGAAG). For the quantitative analysis of PCR products

oligos HT1, HG1, HCD1, HHSP1, RT1, RG1, RCD1, RSP1 or CAD1 were end labeled

with γ32P-ATP. Amplification was performed in a T3 Thermocycler (Biometra®) for 31

cycles (TERT), 19 cycles (GAPDH) and 21 cycles (CAD) at 940C for 45 sec, 600C for 45

sec, and 720C for 60 sec, followed by a final extension step at 720C for 5 min. The optimal

number of cycles for exponential amplification was determined by kinetic analysis. All

quantitative PCR were repeated at least two times. PCR products were resolved on 6%

polyacrylamide gel and quantitated using Molecular Dynamics Phosphor Imaging System.

Immunoprecipitation

HEK 293 cells were transfected by the calcium phosphate method. Cells were lysed using F

buffer (Sommer et al. 1998) 48 hours after transfection. For immunoprecipitations, lysates

were incubated with anti-FLAG antibodies with constant rocking at 40C overnight,

followed by several 1 hour washes in F buffer and/or F buffer supplemented with 0.3 M

NaCl. Elution was carried out overnight at 40C in F buffer supplemented with FLAG

134

peptide (100 µg/ml). To determine levels of protein production, proteins on or eluted

from beads were resolved on SDS polyacrylamide gels and Western-blotted to a

nitrocellulose membrane. Membranes were probed with anti-Flag antibodies (M2, Sigma),

anti-c-Myc (N262), anti-N-myc (C19), anti-TRRAP (T17), and anti-GCN5 (N18), all from

Santa Cruz. Antibodies against TIP49 and BAF53 were previously described (Park et al.

2002) (Wood et al. 2000).

Retrovirally infected human lung fibroblasts (IMR90) were cross-linked by adding

formaldehyde (Fisher Scientific, final concentration 1%) directly to the cells on a rocking

tissue culture dish for 10 min at room temperature. Cross-linking was terminated by

addition of glycine to a final concentration of 125 mM. Fixed cells were washed with PBS

containing 0.5 mM PMSF followed by scraping cells in swelling buffer (5 mM Pipes (pH

8.0), 85 mM KCl, 0.5% NP40, 0.5 mM PMSF, and 100 ng/ml leupeptin and aprotinin).

Cells were incubated on ice for 20 min and then dounce homogenized. Nuclei were

harvested by centrifugation (4500 g) following resuspension in sonication buffer (0.1%

SDS, F-buffer, 0.5 mM PMSF, 100 ng/ml leupeptin and aprotinin) and sonicated on ice to

obtain 1,000-2,000 bp long DNA fragments. After sonication, lysates were cleared by

microcentrifugation at 14,000 rpm. Lysates were normalized according to their OD

measurements. Each lysate was diluted six times with dilution buffer (F-buffer, 0.5 mM

PMSF, 100 ng/ml leupeptin and aprotinin) and normalized by dilution in IP buffer (0.015%

SDS, F-buffer 0.5 mM PMSF, and 100 ng/ml leupeptin and aprotinin). The normalized

chromatin lysates were precleared overnight with pre-blocked protein A/G beads. Protein

A/G beads were pre-blocked by incubation in IP buffer containing 1ug/ml salmon sperm

DNA and 1µg/ml BSA for three hours at 4°C. Precleared lysates were incubated with 1.5

135

µg or 5 µL of anti-acetylated histone H3 or H4 antibodies (Upstate Biotechnology)

respectively, at 4°C overnight. Beads were washed and eluted, followed by reversing

cross-linking and DNA isolation using the Qiagen "PCR purification kit". Eluted material

was subjected to quantitative PCR amplification. PCR was performed using the following

sets of primers: human TERT promoter specific primers: SC-9

(AGTGGATTCGCGGGCAC-AGA) and SC-10 (AGCACCTCGCGGTAGTGGCT). To

normalize samples by the amount of nonspecific DNA, we amplified a region in the third

intron of human β-globin gene using the following primers: SC-46 (ATCTTCC-

TCCCACAGCTCCT) and SC-47 (TTTGCAGCCTCACCTTCTTT). Oligonucleotides

SC-9 and SC-46 were end-labeled with γ32P-ATP. The optimal number of cycles for

exponential amplification was determined by kinetic analysis. PCR products were resolved

on a 4% or 6% polyacrylamide gel. PCR products were quantitated and normalized using

the Molecular Dynamics Phosphor Imaging System.

HAT Assay

Complexes containing FLAG-tagged proteins were eluted from the FLAG-conjugated

beads by addition of excess FLAG-peptide (100 µg/ml) in HAT assay buffer (50 mM Tris-

HCL at pH 8.0, 10% glycerol, 50 mM KCL, 0.1 mM EDTA, 1 mM DTT, 1 mM PMSF, 10

mM butyric acid). Core histones from HeLa cells (3 µg) along with 14C-acetyl-CoA were

incubated with each eluate at 300 C for 30 minutes.

Acknowledgements We are grateful to Heather Van Buskirk for the critical reading of the manuscript. We

thank H. Land and B. Amati for communicating experiments prior to publication. This

136

work was supported by a grant from the New Jersey Commission on Cancer Research to

SC and by grants from the NIH to MDC.

137

REFERENCES

Acker B, Hoppe RT, Colby TV, Cox RS, Kaplan HS et al. (1983) Histologic conversion

in the non-Hodgkin's lymphomas. J Clin Oncol 1(1): 11-16. Alberts AS, Geneste O, Treisman R (1998) Activation of SRF-regulated chromosomal

templates by Rho-family GTPases requires a signal that also induces H4 hyperacetylation. Cell 92(4): 475-487.

Artandi SE, DePinho RA (2000) A critical role for telomeres in suppressing and facilitating carcinogenesis. Curr Opin Genet Dev 10: 39-46.

Ashburner M, Ball CA, Blake JA, Botstein D, Butler H et al. (2000) Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet 25(1): 25-29.

Askew DS, Ashmun RA, Simmons BC, Cleveland JL (1991) Constitutive c-myc expression in an IL-3-dependent myeloid cell line suppresses cell cycle arrest and accelerates apoptosis. Oncogene 6(10): 1915-1922.

Barbeau D, Marcellus RC, Bacchetti S, Bayley ST, Branton PE (1992) Quantitative analysis of regions of adenovirus E1A products involved in interactions with cellular proteins. Biochem Cell Biol 70: 1123-1134.

Barbeau D, Charbonneau R, Whalen SG, Bayley ST, Branton PE (1994) Functional interactions within adenovirus E1A protein complexes. Oncogene 9: 359-373.

Barrett J, Birrer MJ, Kato GJ, Dosaka-Akita H, Dang CV (1994) Activation domains of L-Myc and c-Myc determine their transforming potencies in rat embryo cells. Mol Cell Biol 12: 3130-3137.

Bello-Fernandez C, Packham G, Cleveland JL (1993) The ornithine decarboxylase gene is a transcriptional target of c-Myc. Proc Natl Acad Sci USA 90: 7804-7808.

Berns K, Martins C, Dannenberg JH, Berns A, te Riele H et al. (2000) A genetic screen to identify genes that rescue the slow growth phenotype of c-myc null fibroblasts. Oncogene 19: 3330-3334.

Birrer MJ, Segal S, Degreve JS, Kaye F, Sausville EA et al. (1988) L-myc cooperates with ras to transform primary rat embryo fibroblasts. Mol Cell Biol 8(6): 2668-2673.

Bissell MJ, Radisky D (2001) Putting tumours in context. Nat Rev Cancer 1(1): 46-54. Blackwell TK, Huang J, Ma A, Kretzner L, Alt FW et al. (1993) Binding of myc proteins

to canonical and noncanonical DNA sequences. Mol Cell Biol 13: 5216-5224. Blackwood EM, Eisenman RN (1991) Max: a helix-loop-helix zipper protein that forms a

sequence-specific DNA-binding complex with Myc. Science 251: 1211-1217. Blasco MA, Lee HW, Hande MP, Samper E, Lansdorp PM et al. (1997) Telomere

shortening and tumor formation by mouse cells lacking telomerase RNA. Cell 91: 25-34.

Bouchard C, Thieke K, Maier A, Saffrich R, Hanley-Hyde J et al. (1999) Direct induction of cyclin D2 by Myc contributes to cell cycle progression and sequestration of p27. EMBO J 18: 5321-5333.

Boyd KE, Farnham PJ (1997) Myc versus USF: discrimination at the cad gene is determined by core promoter elements. Mol Cell Biol 17: 2529-2537.

138

Boyle EI, Weng S, Gollub J, Jin H, Botstein D et al. (2004) GO:TermFinder--open source software for accessing Gene Ontology information and finding significantly

enriched Gene Ontology terms associated with a list of genes. Bioinformatics 20(18): 3710-3715.

Braunstein M, Rose AB, Holmes SG, Allis CD, Broach JR (1993) Transcriptional silencing in yeast is associated with reduced nucleosome acetylation. Genes Dev 7(4): 592-604.

Brough DE, Hofmann TJ, Ellwood KB, Townley RA, Cole MD (1995) An essential domain of the c-Myc protein interacts with a nuclear factor that is also required for E1A-mediated transformation. Mol Cell Biol 15: 1536-1544.

Brown PO, Botstein D (1999) Exploring the new world of the genome with DNA microarrays. Nat Genet 21(1 Suppl): 33-37.

Bush A, Mateyak M, Dugan K, Obaya A, Adachi S et al. (1998) c-myc null cells misregulate cad and gadd45 but not other proposed c-myc targets. Genes Dev 12: 3797-3802.

Cario G, Stanulla M, Fine BM, Teuffel O, Neuhoff NV et al. (2005) Distinct gene expression profiles determine molecular treatment response in childhood acute lymphoblastic leukemia. Blood 105(2): 821-826.

Cawley S, Bekiranov S, Ng HH, Kapranov P, Sekinger EA et al. (2004) Unbiased mapping of transcription factor binding sites along human chromosomes 21 and 22 points to widespread regulation of noncoding RNAs. Cell 116(4): 499-509.

Chang HY, Sneddon JB, Alizadeh AA, Sood R, West RB et al. (2004) Gene expression signature of fibroblast serum response predicts human cancer progression: similarities between tumors and wounds. PLoS Biol 2(2): E7.

Chang HY, Nuyten DS, Sneddon JB, Hastie T, Tibshirani R et al. (2005) Robustness, scalability, and integration of a wound-response gene expression signature in predicting breast cancer survival. Proc Natl Acad Sci U S A 102(10): 3738-3743.

Charron J, Malynn BA, Fisher P, Stewart V, Jeannotte L et al. (1992) Embryonic lethality in mice homozygous for a targeted disruption of the N-myc gene. Genes Dev 6(12A): 2248-2257.

Chung CH, Parker JS, Karaca G, Wu J, Funkhouser WK et al. (2004) Molecular classification of head and neck squamous cell carcinomas using patterns of gene expression. Cancer Cell 5(5): 489-500.

Colby WW, Chen EY, Smith DH, Levinson AD (1983) Identification and nucleotide sequence of a human locus homologous to the v-myc oncogene of avian myelocytomatosis virus MC29. Nature 301(5902): 722-725.

Coller HA, Grandori C, Tamayo P, Colbert T, Lander ES et al. (2000) Expression analysis with oligonucleotide microarrays reveals that MYC regulates genes involved in growth, cell cycle, signaling, and adhesion. Proc Natl Acad Sci USA 97: 3260-3265.

Cooper G (1982) Transforming genes of chicken bursal lymphomas. J Cell Physiol Suppl(1): 209-212.

Coppola JA, Cole MD (1986) Constitutive c-myc oncogene expression blocks mouse erythroleukaemia cell differentiation but not commitment. Nature 320(6064): 760-763.

139

Dang CV (2005) The Myc Target Gene Network and Cellular Proliferation, Growth, Adhesion and Metabolism. submitted.

Dang CV, Resar LM, Emison E, Kim S, Li Q et al. (1999) Function of the c-Myc oncogenic transcription factor. Exp Cell Res 253(1): 63-77.

Davis AC, Wims M, Spotts GD, Hann SR, Bradley A (1993) A null c-myc mutation causes lethality before 10.5 days of gestation in homozygotes and reduced fertility in heterozygous female mice. Genes Dev 7(4): 671-682.

de Alboran IM, O'Hagan RC, Gartner F, Malynn B, Davidson L et al. (2001) Analysis of C-MYC function in normal cells via contitional gene-targeted mutation. Immunity 14: 45-55.

Deleu L, Shellard S, Alevizopoulos K, Amati B, Land H (2001) Recruitment of TRRAP required for oncogenic transformation by E1A. Oncogene 20: 8270-8275.

Eberhardy S, D'Cunha C, Farnham P (2000) Direct examination of histone acetylation on Myc target genes using chromatin immunoprecipitation. J Biol Chem 275: 33798-33805.

Eberhardy SR, Farnham PJ (2001) c-Myc mediates activation of the cad promoter via a post-RNA polymerase II recruitment mechanism. J Biol Chem 276: 48562-48571.

Eberwine J (1996) Amplification of mRNA populations using aRNA generated from immobilized oligo(dT)-T7 primed cDNA. Biotechniques 20(4): 584-591.

Eisen MB, Spellman PT, Brown PO, Botstein D (1998) Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A 95(25): 14863-14868.

Ellwood-Yen K, Graeber TG, Wongvipat J, Iruela-Arispe ML, Zhang J et al. (2003) Myc-driven murine prostate cancer shares molecular features with human prostate tumors. Cancer Cell 4(3): 223-238.

Evan GI, Wyllie AH, Gilbert CS, Littlewood TD, Land H et al. (1992) Induction of apoptosis in fibroblasts by c-myc protein. Cell 69(1): 119-128.

Facchini LM, Penn LZ (1998) The molecular role of Myc in growth and transformation: recent discoveries lead to new insights. FASEB J 12: 633-651.

Felsher DW, Bishop JM (1999a) Transient excess of MYC activity can elicit genomic instability and tumorigenesis. Proc Natl Acad Sci U S A 96(7): 3940-3944.

Felsher DW, Bishop JM (1999b) Reversible tumorigenesis by MYC in hematopoietic lineages. Mol Cell 4(2): 199-207.

Felsher DW, Zetterberg A, Zhu J, Tlsty T, Bishop JM (2000) Overexpression of MYC causes p53-dependent G2 arrest of normal fibroblasts. Proc Natl Acad Sci U S A 97(19): 10544-10548.

Fernandez PC, Frank SR, Wang L, Schroeder M, Liu S et al. (2003) Genomic targets of the human c-Myc protein. Genes Dev 17(9): 1115-1129.

Flint J, Shenk T (1997) Viral transactivating proteins. Ann Rev Genet 31: 177-212. Frank SR, Schroeder M, Fernandez P, Taubert S, Amati B (2001) Binding of c-Myc to

chromatin mediates mitogen-induced acetylation of histone H4 and gene activation. Genes Dev 15: 2069-2082.

Freytag SO, Dang CV, Lee WMF (1990) Definition of the activities and properties of c-myc required to inhibit cell differentiation. Cell Growth Diff 1: 339-343.

140

Fu TF, Rife JP, Schirch V (2001) The role of serine hydroxymethyltransferase isozymes in one-carbon metabolism in MCF-7 cells as determined by 13-C NMR. Arch Biochem

Biophys 393: 42-50. Fuchs M, Gerber J, Drapkin R, Sif S, Ikura T et al. (2001) The p400 complex is an

essential E1A transformation target. Cell 106: 297-307. Gallant P, Shiio Y, Cheng PF, Parkhurst SM, Eisenman RN (1996) Myc and Max

homologs in Drosophila. Science 274: 1523-1527. Garrow TA, Brenner AA, Whitehead VM, Chen XN, Duncan RG et al. (1993) Cloning of

human cDNAs encoding mitochondrial and cytosolic serine hydroxymethyltransferases and chromosomal localization. J Biol Chem 268: 11910-11906.

Girgis S, Nasrallah IM, Suh JR, Oppenheim E, Zanetti KA et al. (1998) Molecular cloning, characterization and alternative splicing of the human cytoplasmic serine hydroxymethyltransferase gene. Gene 210: 315-324.

Greenberg R, O'Hagan R, Deng H, Xiao Q, Hann S et al. (1999) Telomerase reverse transcriptase gene is a direct target of c-Myc but is not functionally equivalent in cellular transformation. Oncogene 18: 1219-1226.

Greider C (1999) Telomerase activation. One step on the road to cancer? Trends Genet 15: 109-112.

Guo QM, Malek RL, Kim S, Chiao C, He M et al. (2000) Identification of c-myc responsive genes using rat cDNA microarray. Cancer Res 60: 5922-5928.

Halees AS, Leyfer D, Weng Z (2003) PromoSer: A large-scale mammalian promoter and transcription start site identification service. Nucleic Acids Res 31(13): 3554-3559.

Hall PA, McKee PH, Menage HD, Dover R, Lane DP (1993) High levels of p53 protein in UV-irradiated normal human skin. Oncogene 8(1): 203-207.

Hannon GJ, Sun P, Carnero A, Xie LY, Maestro R et al. (1999) MaRX: an approach to genetics in mammalian cells. Science 283: 1129-1130.

Harris SL, Levine AJ (2005) The p53 pathway: positive and negative feedback loops. Oncogene 24(17): 2899-2908.

Hatton KS, Mahon K, Chin L, Chiu FC, Lee HW et al. (1996) Expression and activity of L-Myc in normal mouse development. Mol Cell Biol 16: 1794-1804.

Hermeking H, Rago C, Schuhmacher M, Li Q, Barett JF et al. (2000) Identification of CDK4 as a target of c-MYC. Proc Natl Acad Sci USA 97: 2229-2234.

Horning SJ, Rosenberg SA (1984) The natural history of initially untreated low-grade non-Hodgkin's lymphomas. N Engl J Med 311(23): 1471-1475.

Hu Z, Fan C, Marron JS, He X, Qaqish BF et al. (2005) The Molecular Portraits of Breast Tumors Are Conserved Across Microarray Platforms. Submitted.

Ikura T, Ogryzko VV, Grigoriev M, Groisman R, Wang J et al. (2000) Involvement of the TIP60 histone acetylase complex in DNA repair and apoptosis. Cell 102: 463-473.

Iritani BM, Eisenman RN (1999) c-Myc enhances protein synthesis and cell size during B lymphocyte development. Proc Natl Acad Sci USA 96: 13180-13185.

Johnston LA, Prober DA, Edgar BA, Eisenman RN, Gallant P (1999) Drosophila myc regulates cellular growth during development. Cell 98: 779-790.

141

Kastanos EK, Woldman YY, Appling DR (1997) Role of mitochondrial and cytoplasmic serine hydroxymethyltransferase isozymes in de novo purine synthesis in

Saccharomyces cerevisiae. Biochemistry 36: 14956-14964. Keath EJ, Caimi PG, Cole MD (1984) Fibroblast lines expressing activated c-myc

oncogenes are tumorigenic in nude mice and syngeneic animals. Cell 39(2 Pt 1): 339-348.

Kuo MH, Allis CD (1999) In vivo cross-linking and immunoprecipitation for studying dynamic protein:DNA associations in a chromatin environment. Methods 19: 425-433.

Kuttler F, AmeÂ P, Clark H, Haughey C, Mougin C et al. (2001) c-myc box II mutations in Burkitt's lymphoma-derived alleles reduce cell-transformation activity and lower response to broad apoptotic stimuli. Oncogene 20: 6084-6094.

Land H, Parada LF, Weinberg RA (1983) Tumorigenic conversion of primary embryo fibroblasts requires at least 2 cooperating oncogenes. Nature 304(5927): 596-602.

Landay M, Oster SK, Khosravi F, Grove LE, Yin X et al. (2000) Promotion of growth and apoptosis in c-myc nullizygous fibroblasts by other members of the myc oncoprotein family. Cell Death Differ 7: 697-705.

Lang SE, McMahon SB, Cole MD, Hearing P (2001) E2F transcriptional activation requires TRRAP and GCN5 cofactors. J Biol Chem 276: 32627-32634.

Lapointe J, Li C, Higgins JP, van de Rijn M, Bair E et al. (2004) Gene expression profiling identifies clinically relevant subtypes of prostate cancer. Proc Natl Acad Sci U S A 101(3): 811-816.

Lasorella A, Iavarone A, Israel MA (1996) Id2 specifically alters regulation of the cell cycle by tumor suppressor proteins. Mol Cell Biol 16(6): 2570-2578.

Liao DJ, Dickson RB (2000) c-Myc in breast cancer. Endocr Relat Cancer 7(3): 143-164. Littlewood TD, Hancock DC, Danielian PS, Parker MG, Evan GI (1995) A modified

oestrogen receptor ligand-binding domain as an improved switch for the regulation of heterologous proteins. Nucl Acids Res 23(10): 1686-1690.

Lockhart DJ, Dong H, Byrne MC, Follettie MT, Gallo MV et al. (1996) Expression monitoring by hybridization to high-density oligonucleotide arrays. Nat Biotechnol 14(13): 1675-1680.

Lossos IS, Alizadeh AA, Diehn M, Warnke R, Thorstenson Y et al. (2002) Transformation of follicular lymphoma to diffuse large-cell lymphoma: alternative patterns with increased or decreased expression of c-myc and its regulated genes. Proc Natl Acad Sci U S A 99(13): 8886-8891.

Malynn BA, de Alboran IM, O'Hagan RC, Bronson R, Davidson L et al. (2000) N-myc can functionally replace c-myc in murine development, cellular growth, and differentiation. Genes Dev 14(11): 1390-1399.

Mao DY, Barsyte-Lovejoy D, Ho CS, Watson JD, Stojanova A et al. (2004) Promoter-binding and repression of PDGFRB by c-Myc are separable activities. Nucleic Acids Res 32(11): 3462-3468.

Marinkovic D, Marinkovic T, Kokai E, Barth T, Moller P et al. (2004) Identification of novel Myc target genes with a potential role in lymphomagenesis. Nucleic Acids Res 32(18): 5368-5378.

142

Mateyak MK, Obaya AJ, Adachi S, Sedivy JM (1997) Phenotypes of Myc-deficient rat fibroblasts isolated by targeted homologous recombination. Cell Growth Differ 8: 1039-

1048. McMahon SB, Wood MA, Cole MD (2000) The c-Myc cofactor TRRAP recruits the

histone acetylase hGCN5. Mol Cell Biol 20: 556-562. McMahon SB, Van Buskirk HA, Dugan KA, Copeland TD, Cole MD (1998) The novel

ATM-related protein TRRAP is an essential cofactor for the c-Myc and E2F oncoproteins. Cell 94: 363-374.

Menssen A, Hermeking H (2002) Characterization of the c-MYC-regulated transcriptome by SAGE: identification and analysis of c-MYC target genes. Proc Natl Acad Sci U S A 99(9): 6274-6279.

Miltenberger RJ, Sukow KA, Farnham PJ (1995) An E-Box-mediated increase in cad transcription at the G1/S-phase boundary is suppressed by inhibitory c-Myc mutants. Mol Cell Biol 15(5): 2527-2535.

Nesbit CE, Tersak JM, Prochownik EV (1999) MYC oncogenes and human neoplastic disease. Oncogene 18(19): 3004-3016.

Nikiforov MA, Kotenko I, Petrenko O, Beavis A, Valenick L et al. (2000) Complementation of Myc-dependent cell proliferation by cDNA expression library screening. Oncogene 19: 4828-4831.

Nikiforov MA, Chandriani S, O'Connell B, Petrenko O, Kotenko I et al. (2002a) A functional screen for Myc-responsive genes reveals serine hydroxymethyltransferase, a major source of one-carbon unit for cell metabolism. Mol Cell Biol 22: 5793-5800.

Nikiforov MA, Chandriani S, Park J, Kotenko I, Matheos D et al. (2002b) TRRAP-dependent and TRRAP-independent transcriptional activation by Myc family oncoproteins. Mol Cell Biol 22: 5054-5063.

O'Connell BC, Cheung AF, Simkevich CP, Tam W, Ren X et al. (2003) A large scale genetic analysis of c-Myc-regulated gene expression patterns. J Biol Chem.

O'Hagan RC, Ohh M, David G, de Alboran IM, Alt FW et al. (2000) Myc-enhanced expression of Cul1 promotes ubiquitin-dependent proteolysis and cell cycle progression. Genes Dev 14: 2185-2191.

Ogryzko VV, Kotani T, Zhang X, Schiltz RL, Howard T et al. (1998) Histone-like TAFs within the PCAF histone acetylase complex. Cell 94: 35-44.

Orian A, Van Steensel B, Delrow J, Bussemaker HJ, Li L et al. (2003) Genomic binding by the Drosophila Myc, Max, Mad/Mnt transcription factor network. Genes Dev 17(9): 1101-1114.

Park J, Wood MA, Cole MD (2002) BAF53 forms distinct nuclear complexes and functions as a critical c-Myc-interacting nuclear cofactor for oncogenic transformation. Mol Cell Biol 22: 1307-1316.

Park J, Kunjibettu S, McMahon SB, Cole MD (2001) The ATM-related domain of TRRAP is required for histone acetyltransferase recruitment and Myc-dependent oncogenesis. Genes Dev 15: 1619-1624.

Pear WS, Nolan GP, Scott ML, Baltimore D (1993) Production of high-titer helper-free retroviruses by transient transfection. Proc Natl Acad Sci U S A 90: 8392-8396.

Pelengaris S, Khan M, Evan G (2002) c-MYC: more than just a matter of life and death. Nat Rev Cancer 2(10): 764-776.

143

Pelengaris S, Littlewood T, Khan M, Elia G, Evan G (1999) Reversible activation of c-Myc in skin: induction of a complex neoplastic phenotype by a single oncogenic lesion.

Mol Cell 3(5): 565-577. Petrenko O, Beavis A, Klaine M, Kittappa R, Godin I et al. (1999) The molecular

characterization of the fetal stem cell marker AA4. Immunity 10: 691-700. Pfendner W, Pizer LI (1980) The metabolism of serine and glycine in mutant lines of

Chinese hamster ovary cells. Arch Biochem Biophys 200: 503-512. Quackenbush J (2002) Microarray data normalization and transformation. Nat Genet 32

Suppl: 496-501. Ralston R (1991) Complementation of transforming domains of E1a/myc chimaeras.

Nature 353: 866-868. Rao RN, Stamm NB, Otto K, Kovacevic S, Watkins SA et al. (1999) Conditional

transformation of rat embryo fibroblast cells by a cyclin D1-cdk4 fusion gene. Oncogene 18: 6343-6356.

Reichel R, Kovesdi I, Nevins JR (1987) Developmental control of a promoter-specific factor that is also regulated by the E1A gene product. Cell 13: 501-506.

Sawai SA, Shimono K, Yakamatsu Y, Palmes C, Hanaoka K et al. (1993) Defects of embryonic organogenesis resulting from targeted disruption of the N-myc gene in the mouse. Development 117: 1445-1455.

Schuhmacher M, Kohlhuber F, Holzel M, Kaiser C, Burtscher H et al. (2001) The transcriptional program of a human B cell line in response to Myc. Nuc Acids Res 29: 397-406.

Sheiness D, Bishop JM (1979) DNA and RNA from uninfected vertebrate cells contain nucleotide sequences related to the putative transforming gene of avian myelocytomatosis virus. J Virol 31(2): 514-521.

Shim H, Dolde C, Lewis BC, Wu CS, Dang G et al. (1997) c-Myc transactivation of LDH-A: implications for tumor metabolism and growth. Proc Natl Acad Sci USA 94: 6658-6663.

Shyamsundar R, Kim YH, Higgins JP, Montgomery K, Jorden M et al. (2005) A DNA microarray survey of gene expression in normal human tissues. Genome Biol 6(3): R22.

Snell K (1980) Liver enzymes of serine metabolism during neonatal development of the rat. Biochem J 190: 451-455.

Snell K (1985) Enzymes of serine metabolism in normal and neoplastic rat tissues. Biochim Biophys Acta 843: 276-281.

Snell K, Baumann U, Byrne PC, Chave KJ, Renwick SB et al. (2000) The genetic organization and protein crystallographic structure of human serine hydroxymethyltransferase. Adv Enzyme Regul 40: 353-403.

Sommer A, Bousset K, Kremmer E, Austen M, Luscher B (1998) Identification and characterization of specific DNA-binding complexes containing members of the Myc/Max/Mad network of transcriptional regulators. J Biol Chem 273: 6632-6642.

Soucek L, Evan G (2002) Myc-Is this the oncogene from Hell? Cancer Cell 1(5): 406-408.

144

Stanton BR, Perkins AS, Tessarollo L, Sassoon DA, Parada LF (1992) Loss of N-myc function results in embryonic lethality and failure of the epithelial component of the

embryo to develop. Genes Dev 6: 2235-2247. Stone J, De Lange T, Ramsay G, Jakobovits E, Bishop JM et al. (1987) Definition of

regions in human c-myc that are involved in transformation and nuclear localization. Mol Cell Biol 7(5): 1697-1709.

Stover PJ, Chen LH, Suh JR, Stover DM, Keyomarsi K et al. (1997) Molecular cloning, characterization, and regulation of the human mitochondrial serine hydroxymethyltransferase gene. J Biochem Chem 272(1842-1848).

Thorndike J, Pelliniemi TT, Beck WS (1979) Serine hydroxymethyltransferase activity and serine incorporation in leukocytes. Cancer Res 39: 3435-3440.

Trumpp A, Refaell Y, Oskarsson T, Gasser S, Murphy M et al. (2001) c-Myc regulates mammalian body size by controlling cell number but not cell size. Nature 414: 768-773.

Tusher VG, Tibshirani R, Chu G (2001) Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 98(9): 5116-5121.

Valsesia-Wittmann S, Magdeleine M, Dupasquier S, Garin E, Jallas AC et al. (2004) Oncogenic cooperation between H-Twist and N-Myc overrides failsafe programs in cancer cells. Cancer Cell 6(6): 625-630.

van 't Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA et al. (2002) Gene expression profiling predicts clinical outcome of breast cancer. Nature 415(6871): 530-536.

Vassilev A, Yamauchi J, Kotani T, Prives C, Avantaggiati ML et al. (1998) The 400 kDa subunit of the PCAF histone acetylase complex belongs to the ATM superfamily. Mol Cell 2(6): 869-875.

Vennstrom B, Sheiness D, Zabielski J, Bishop JM (1982) Isolation and characterization of c-myc, a cellular homolog of the oncogene (v-myc) of avian myelocytomatosis virus strain 29. J Virol 42(3): 773-779.

Wagner AJ, Meyers C, Laimins L, Hay N (1993) c-Myc induces the expression and activity of ornithine decarboxylase. Cell Growth Differ 4(11): 879-883.

Wang HY, Malek RL, Kwitek AE, Greene AS, Luu TV et al. (2003) Assessing unmodified 70-mer oligonucleotide probe performance on glass-slide microarrays. Genome Biol 4(1): R5.

Wang J, Xie L, Allan S, Beach D, Hannon G (1998) Myc activates telomerase. Genes Dev 12: 1769-1774.

Watson JD, Oster SK, Shago M, Khosravi F, Penn LZ (2002) Identifying genes regulated in a Myc-dependent manner. J Biol Chem 277(40): 36921-36930.

Watt R, Stanton LW, Marcu KB, Gallo RC, Croce CM et al. (1983) Nucleotide sequence of cloned cDNA of human c-myc oncogene. Nature 303(5919): 725-728.

Wood MA, McMahon SB, Cole MD (2000) An ATPase/helicase complex is an essential cofactor for oncogenic transformation by c-Myc. Mol Cell 5: 321-330.

Wu KJ, Polack A, Dalla-Favera R (1999) Coordinated regulation of iron-controlling genes, H-ferritin and IRP2, by c-MYC. Science 283: 676-679.

Xiao Q, Claassen G, Shi J, Adachi S, Sedivy J et al. (1998) Transactivation-defective c-MycS retains the ability to regulate proliferation and apoptosis. Genes Dev 12(24): 3803-3808.

145

Xu D, Popov N, Hou M, Wang Q, Bjorkholm M et al. (2001) Switch from Myc/Max to Mad1/Max binding and decrease in histone acetylation at the telomerase reverse

transcriptase promoter during differentiation of HL60 cells. Proc Natl Acad Sci USA 98: 3826-3831.

Yano T, Jaffe ES, Longo DL, Raffeld M (1992) MYC rearrangements in histologically progressed follicular lymphomas. Blood 80(3): 758-767.

Zeller KI, Jegga AG, Aronow BJ, O'Donnell KA, Dang CV (2003) An integrated database of genes responsive to the Myc oncogenic transcription factor: identification of direct genomic targets. Genome Biol 4(10): R69.

146

myc mediated transcriptional modulation sanjay j....

Documents