myc mediated transcriptional modulation sanjay j....
TRANSCRIPT
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
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 ∆
15’ 30’ 1hr 2hrs 4hrs 8hrs 12hrs 24hrs 36hrs 48hrs 0hr
- + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - +- + 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
- + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - + - +- + 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:
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
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 ∆MbIIA.
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
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
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
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
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
morphogenesis 5.1E-03
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
morphogenesis 3.1E-05
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
MycER MycER∆MBIIVectorVector
WT∆MBII
repressed
induced
MycER MycER∆MBIIVectorVector
WT∆MBII
repressed
induced
MycER MycER∆MBIIVectorVector
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
Materials and Methods
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∆
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
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
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
Materials and Methods
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, Ame 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