University of Southern California Dissertations and Theses

Runx2 interactions with the osteoblast genome

(USC Thesis Other)

Runx2 interactions with the osteoblast genome

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content RUNX2 INTERACTIONS WITH THE OSTEOBLAST GENOME

by

Steven Pregizer

A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BIOCHEMISTRY & MOLECULAR BIOLOGY)

August 2008

Copyright 2008 Steven Pregizer
ii
Dedication

This work is dedicated primarily to my teachers. In particular, I would like to
acknowledge Dr. Baruch Frenkel, a teacher among teachers, for his patient guidance
during my time in his lab. Additionally, I would like to acknowledge the faculty of
the Department of Biochemistry & Molecular Biology at USC for their commitment
to graduate level education. This work is also dedicated to my friends, family, and to
my wonderful wife, Teresa. Every accomplishment of mine has been due, in no
small part, to their support and encouragement.

iii
Table of Contents

Dedication ii
List of Tables vi
List of Figures vii
Abstract x
Chapter 1: Introduction 1
Importance of Runx2 in Osteoblast Biology 2
Molecular mechanims mediating the important role of Runx2 5
are poorly understood

Chapter 2: Materials and Methods 10
Cell Culture 10
Runx2 Over-expression 10
Statistical Analysis 11
RNA Analysis 11
Protein Analysis 11
ChIP Assays 12
ChIP-chip 12
ChIP Display 12

Chapter 3: Identification of Transcription Factor Target Genes 14
by ChIP Display (CD)
3.1 Introduction 14
3.2 Materials 19
Digestion and Linker Ligation 19
PCR Amplification 19
PAGE 19
Target Identification 20
3.3 Methods 21
Digestion and Linker Ligation 22
PCR Amplification 23
PAGE and Target Isolation 23
Target Identification 26
3.4 Notes 28

iv
Chapter 4: Identification of Novel Runx2 Targets in Osteoblasts: 39
Cell Type-Specific BMP-Dependent Regulation of Tram2
4.1 Introduction 39
4.2 Results 42
ChIP Display discloses novel Runx2 genomic targets 42
in living osteoblasts
Runx2 inhibits Tram2 expression in non-osteoblasts 45
Tram2 is positively regulated along with osteocalcin 47
late during BMP-2-induced osteoblastic differentiation
of C3H10T1/2 mouse embryonic fibroblasts
Tram2 is up-regulated soon after BMP-2 treatment 50
of ST2 bone marrow-derived stromal cells
Runx2, osteocalcin and Tram2 are transiently 52
upregulated in BMP-2-treated C2C12 myoblasts
Tram2 expression in MC3T3-E1 osteoblastic cells 54
is regulated by BMP-2 in a developmental stage-
specific manner
4.3 Discussion 56
Cell Type-dependent and Gene-Specific 57
Transcriptional Regulation by Runx2
BMP-Dependent Transcriptional Regulation by Runx2 58
Potential Novel Mediators of Osteoblast Development 59
and Function

Chapter 5: An Expression-Based Approach for Identifying 62
Novel Runx2 Targets

Chapter 6: Optimzation of Critical Parameters for Runx2 82
Location Analysis in MC3T3-E1 Pre-Osteoblasts.

Chapter 7: Progressive Recruitment of Runx2 to Genomic Targets 116
Despite Decreasing Expression during Osteoblast Differentiation
7.1 Introduction 116
7.2 Results 117
Paradoxical inverse relationship between Runx2 and 117
osteocalcin expression during osteoblast differentiation
Developmental up-regulation of osteocalcin is attributable 121
to enhanced Runx2 occupancy
Developmental up-regulation of Runx2 occupancy 123
at the novel target, Glt28d2
7.4 Discussion 124

v
Chapter 8: Discussion 128

Genome-wide location analysis of Runx2 in osteoblasts 128
Multiple levels of regulation result in poor correlations 131
between Runx2 expression and activity

References 135

vi
List of Tables

Table I: Oligonucleotides used in Chapter 4 13
Table II: Oligonucleotides used in Chapter 7 13
Table III: Runx2-occupied genomic sites disclosed in the present study 43
Table IV: Genes with altered expression in the presence of exogenous 71
Runx2

Table V: Peaks Identified by Replicate ChIP-chip Experiments 122

vii
List of Figures

Figure 1: Principles of ChIP Display (CD) 18

Figure 2: Target Display and Identification 26

Figure 3: Effect of Dilution and Annealing Temperature on CD 35

Figure 4: Representative ChIP Display results 42

Figure 5: Runx2 interacts with Tram2 45

Figure 6: Runx2 inhibits Tram2 expression in non-osteoblasts 47

Figure 7: Upregulation of Tram2 mRNA in BMP-2 treated 49
embryonic fibroblasts

Figure 8: Early upregulation of Tram2 mRNA in rhBMP-2 51
treated bone marrow-derived stromal cells

Figure 9: Transient upregulation of Tram2 mRNA in rhBMP-2 53
treated C2C12 pre-myoblasts

Figure 10: Tram2 expression in osteoblastic cells is regulated 55
by BMP-2 in a developmental stage-specific manner

Figure 11: Effect of exogenous Runx2 on gene expression in 63
MC3T3-E1 cells

Figure 12: Gel analysis of chromatin isolated from C3H10T1/2 cells 65

Figure 13: ChIP analysis of Runx2 occupancy in C3H10T1/2 cells 66

Figure 14: Gel and ChIP analysis of chromatin collected from 67
C3H10T1/2 cells

Figure 15: Gel analysis of RNA collected from C3H10T1/2 cells 68

Figure 16: RT-qPCR analysis of RNA collected from C3H10T1/2 70
cells transfected with Runx2

viii
Figure 17: RT-qPCR analysis of RNA collected from C3H10T1/2 77
cells transfected with Runx2

Figure 18: RT-qPCR analysis of IL-6 expression in C3H10T1/2 78
cells transfected with different doses of Runx2

Figure 19: RT-qPCR analysis of gene expression in C3H10T1/2 79
cells transfected with Runx2

Figure 20: RT-qPCR analysis of gene expression in C3H10T1/2 81
cells transfected with Runx2

Figure 21: qPCR analysis of Runx2 ChIP before and after amplification 83

Figure 22: qPCR analysis of input DNA before and after amplification 84

Figure 23: qPCR analysis of input DNA before and after amplification 86

Figure 24: qPCR analysis of input DNA before and after amplification 87

Figure 25: qPCR analysis of input DNA before and after amplification 88

Figure 26: qPCR analysis of input DNA before and after amplification 89

Figure 27: Gel analysis of sonicated MC3T3-E1 chromatin 90
before and after amplification

Figure 28: Gel analysis of sonicated MC3T3-E1 chromatin 92

Figure 29: Gel analysis of sonicated MC3T3-E1 chromatin 94

Figure 30: Gel analysis of sonicated MC3T3-E1 chromatin 95

Figure 31: qPCR analysis of Runx2 ChIP performed under 96
various conditions

Figure 32: Gel and ChIP analysis of cross-linked chromatin 97
isolated from MC3T3-E1 cells

Figure 33: Gel and ChIP analysis of cross-linked chromatin 98
isolated from MC3T3-E1 cells

ix
Figure 34: Gel and ChIP analysis of cross-linked chromatin 100
isolated from MC3T3-E1 cells

Figure 35: Gel and ChIP analysis of cross-linked chromatin 102
isolated from MC3T3-E1 cells

Figure 36: Gel and ChIP analysis of cross-linked chromatin 103
isolated from MC3T3-E1 cells

Figure 37: Gel analysis of chromatin isolated from MC3T3-E1 cells 106

Figure 38: ChIP analysis of chromatin isolated from MC3T3-E1 cells 107

Figure 39: Gel analysis of chromatin isolated from MC3T3-E1 cells 109

Figure 40: ChIP analysis of chromatin isolated from MC3T3-E1 cells 110

Figure 41: Gel analysis of chromatin isolated from MC3T3-E1 cells 112

Figure 42: ChIP analysis of chromatin isolated from MC3T3-E1 cells 113

Figure 42: ChIP analysis of chromatin isolated from MC3T3-E1 cells 114

Figure 44: Gel and qPCR analysis of amplified ChIP material 115

Figure 45: Increased Runx2 occupancy, not expression or DNA- 120
binding activity parallels OC up-regulation during osteoblast maturation

Figure 46: Conventional ChIP confirmation of Runx2 binding sites 122
discovered by ChIP-chip

Figure 47: Identification of a novel Runx2-occupied genomic target 124
and characterization of its occupancy during osteoblast differentiation
x
Abstract

Runx2 is a master transcription factor in osteoblasts, yet its mechanism is poorly
understood. In particular, there is a paucity of information about its target genes and
their regulation. To address this, we first used ChIP Display to discover novel
genomic targets occupied by Runx2 in living MC3T3-E1 osteoblastic cells. One of
these targets was located within the promoter of Tram2, whose product facilitates
proper folding of type I collagen. We demonstrated that Tram2 mRNA levels were
altered by exogenous Runx2, and that this occurred in a BMP- and cell type-
dependent manner. Thus, Tram2 is likely a Runx2 target gene and may participate in
its osteogenic function. Next, we measured endogenous Runx2 expression in
MC3T3-E1 cells during development of the osteoblast phenotype, to see if it could
explain the dramatic increase in mRNA levels of Osteocalcin (OC), a classic Runx2
target gene. Surprisingly, we discovered that it could not, as Runx2 expression
decreased over time, along with in vitro DNA binding activity. Instead,
developmental stimulation of OC by Runx2 is attributable to enhanced promoter
occupancy in vivo. A remarkably similar pattern of recruitment was observed at the
Glt28d2 promoter, a novel Runx2 genomic target discovered by ChIP-Chip analysis
of cells in which the OC promoter is maximally occupied. Thus, Runx2 acquires the
ability to access target genes relatively late during development of the osteoblast
phenotype, and this is most likely due to activation of collaborating factors or to
post-translational modification of Runx2 itself. Expanding our knowledge of Runx2
xi
target genes and their regulation is warranted to better understand the regulation of
osteoblast function and to provide opportunities for the development of new bone
anabolics.
1
Chapter 1: Introduction

Runx2 is a member of the Runt-related family of mammalian transcription factors,
which also includes Runx1 and Runx3. Each of these contains a highly-conserved
runt-related domain, so-called because it is homologous to the Drosophila Runt
protein. In mammals, Runx2 was first discovered in the context of viral enhancers.
One lab identified a polyoma virus enhancer binding protein in transformed mouse
fibroblasts, and called it Pebp2 αA [Ogawa et al., 1993]. Simultaneously, another lab
identified a Moloney murine leukemia virus enhancer core-binding factor in thymic
cells and called it Cbfa1 [Wang et al., 1993]. It is now understood that both proteins
were, in fact Runx2. Originally, Runx2 was thought to play a role in modulating the
expression of T-cell specific genes [Satake et al., 1995; Wang et al., 1993]. Then,
reports began to surface from other labs indicating that a Runt-related protein had
been detected in osteoblasts, where it was purported to play a role in the
transcriptional stimulation of osteoblast-specific genes [Geoffroy et al., 1995;
Merriman et al., 1995]. It soon became clear that this molecule was an osteoblast-
specific isoform of the originally-described Pebp2 αA/Cbfa1 that could indeed
regulate transcription of osteoblast-specific genes [Banerjee et al., 1997; Ducy et al.,
1997; Stewart et al., 1997].

2
Importance of Runx2 in Osteoblast Biology
Despite the discovery of its presence in osteoblasts, the importance of Runx2 in
osteoblast biology only began to be fully appreciated when genetic evidence from
knock-out mice became available. Remarkably, homozyogous deletion of Runx2
caused a complete blockage of intramembranous and endochondral ossification,
which resulted in asphyxiation shortly after birth due to an inability of the unossified
ribcage to support breathing [Komori et al., 1997; Otto et al., 1997]. These skeletal
defects were due to the absence of osteoblasts, an observation which led to the
hypothesis that Runx2 is required for osteoblast differentiation. A myriad of in vitro
experiments have both supported and refined this hypothesis. For example, calvarial
cells cultured from Runx2
-/-
mice differentiate more readily into adipocytes or
chondrocytes [Kobayashi et al., 2000; Liu et al., 2007], and have enhanced
proliferative ability [Pratap et al., 2003], suggesting that Runx2 normally functions
in mesenchymal stem cells (MSCs) to support exit from the cell cycle and
commitment to the osteoblast lineage, while blocking commitment to other MSC-
derived lineages. Further support for this notion comes from cell culture
experiments which have shown that exogenous Runx2 is inhibitory for cell cycle
progression [Galindo et al., 2005], and that it can induce the osteoblast phenotype in
cells that either resemble or are derived from MSCs, including primary myoblasts
[Gersbach et al., 2004; Gersbach et al., 2004; Gersbach et al., 2006], primary
osteoprogenitors [Kojima and Uemura, 2005], bone marrow stem cells [Zheng et al.,
2004], bone marrow stromal cells [Byers and Garcia, 2004; Byers et al., 2004; Zhao
3
et al., 2005], adipose tissue-derived stem cells [Zhang et al., 2006], primary dermal
cells [Phillips et al., 2007] and immortalized fibroblasts [Byers et al., 2002].

The importance of Runx2 in osteoblast biology goes beyond the initial level of
commitment to the osteoblast lineage. At first it was thought to promote the function
of osteoblasts post-natally [Ducy et al., 1999]; however, this notion has been turned
almost completely on its head. It began with several reports of osteopenia due to
high bone turnover in transgenic mice over-expressing Runx2 post-natally [Geoffroy
et al., 2002; Liu et al., 2001]. These phenomena went unexplained for some time,
and more reports surfaced of Runx2 transgenic mice with either high turnover or
osteopenic bone phenotypes [Kanatani et al., 2006; Selvamurugan et al., 2006].
Finally, one group proposed a hypothesis to account for the inhibitory role Runx2
seemed to play in bone development. They proposed that Runx2 inhibits the
terminal differentiation of osteoblasts, which then secrete an immature form of bone
that is less resistant to resorption by osteoclasts [Maruyama et al., 2007]. In
accordance with this hypothesis, they showed that transgenic mice expressing a
dominant-negative version of Runx2 post-natally developed a high bone mass
phenotype which was attributable to low bone turnover. Moreover, they showed that
endogenous Runx2 in wild-type mice is strongly expressed in immature osteoblasts,
but is downregulated during osteoblast maturation. Thus, Runx2 is required only for
the commitment of mesenchymal cells to the osteoblast lineage. After this, it
becomes inhibitory for further development of the osteoblast phenotype.
4

The importance of Runx2 in osteoblast biology is further seen in that skeletal
phenotypes are sensitive to small dose changes. For example, heterozygous
disruption of the “bone-specific” type II Runx2 isoform results in osteopenia that
persists to adulthood, whereas homozygous disruption results in early blockage of
endochondral bone formation that increases the severity of osteopenia in adulthood
[Xiao et al., 2005; Xiao et al., 2004]. Moreover, intramembranous bone formation is
impaired in aged Runx2 heterozygous adult mice [Juttner and Perry, 2007; Tsuji et
al., 2004], and they lose bone more rapidly under conditions of mechanical
unloading [Salingcarnboriboon et al., 2006]. Heterozygous deletion of Runx2 also
results in a skeletal phenotype closely resembling that of the mouse model for
cleidocranial dysplasia (CCD), a human disease characterized by skeletal
abnormalities such as patent fontanels, late closure of cranial sutures with Wormian
bones, late erupting secondary dentition, rudimentary clavicles, and short stature
[Otto et al., 1997]. In humans, several types of heterozygous Runx2 mutations are
associated with CCD, including chromosomal translocations, deletions, insertions,
nonsense, and missense mutations [Otto et al., 2002]. Of these, missense mutations
are most frequently reported, the majority of which are found in the runt domain and
affect binding to CBF , resulting in hypomorphic alleles [Matheny et al., 2007]. As
with mutations, polymorphisms in Runx2 are strongly associated with differences in
various skeletal parameters. For example, 3 SNPs in introns 2 and 3 are associated
with variations in femoral length among a Chuvasha population [Ermakov et al.,
5
2006]. Additionally, a synonymous alanine codon polymorphism with alleles GCA
and GCG (noted as A and G alleles, respectively) is found within the glutamine-
alanine repeat (exon 2) of Runx2. The A allele is positively associated with higher
bone mineral density at the femoral neck within a Scottish population, and this
association increases with increasing weight [Vaughan et al., 2004]. It is also
positively associated with increased bone mineral density at all sites tested and is
protective against a common form of osteoporotic fracture in an Australian
population [Vaughan et al., 2002]. Moreover, it is in strong linkage disequilibrium
with the minor alleles of 3 SNPs found in the P2 promoter. These alleles are
positively associated with increased bone mineral density in an Australian
population, and result in an increased activity of the P2 promoter [Doecke et al.,
2006]. Thus, there is a strong association between genetic changes in Runx2 and
bone-related phenotypes, both in mice and humans.

Molecular mechanims mediating the important role of Runx2 are poorly understood
Given the powerful, yet complex role of Runx2 in osteoblast biology, a proper
understanding of its mechanism is warranted. It was appreciated early on that Runx2
is a transcriptional regulator, and that it can influence the expression of many genes
that are expressed in osteoblasts, inlcuding osteocalcin [Ducy et al., 1997], Type I
Collagen [Kern et al., 2001; Tsuji et al., 1998], osteopontin [Kawahata et al., 2003;
Sato et al., 1998], BSP [Javed et al., 2001], collagenase 3 [Jimenez et al., 1999;
Selvamurugan et al., 2000], TGF- βRI [Ji et al., 1998], BMP-2 [Choi et al., 2005],
6
galectin-3 [Stock et al., 2003], Dentin matrix protein 1 [Fen et al., 2002],
Craniosynostosis-associated gene nell-1 [Truong et al., 2007], Smad6 [Wang et al.,
2007], SOST [Sevetson et al., 2004], pituitary tumor transforming 1 protein
interacting protein [Stock et al., 2004], osterix [Nishio et al., 2006], Er α [Lambertini
et al., 2007], ODF/OPGL [Gao et al., 1998], RANKL [Enomoto et al., 2003], OPG
[Thirunavukkarasu et al., 2000], p21
CIP/WAF1
[Westendorf et al., 2002], GNAS, SelM,
elF-4AI, RPS24, CD99/MIC2 [Bertaux et al., 2006; Bertaux et al., 2005], and genes
coding for ribosomal RNA [Young et al., 2007]. Although it is generally considered
a transcriptional activator, Runx2 can also suppress gene expression, depending on
cell and promoter context [Javed et al., 2001].

Given its ability to regulate gene expression in either direction, it follows that Runx2
probably supports differentiation of osteoblasts from their mesenchymal precursors
by activating genes required for the osteoblast phenotype. Moreover, it likely
suppresses genes that are specific to MSCs and other MSC-derived cell types. In this
way, it acts as a ‘master switch’ in osteoblast differentiation, analogous to the role
played by Ppar γ in adipocytes or MyoD in myoblasts. Finally, it may also suppress
genes that are important for terminal osteoblast differentiation, consistent with the
inhibitory role of Runx2 in this process. Although this hypothesis seems obvious, it
has yet to be confirmed experimentally. This is true in spite of the fact that knock-
out mice are available for several known Runx2 targets, none of which appear to
have compromised osteoblast function [Aubin et al., 1996; Bucay et al., 1998;
7
Colnot et al., 1998; Ducy et al., 1996; Inada et al., 2004; Mizuno et al., 1998; Rittling
et al., 1998; Stickens et al., 2004]. One possible explanation for these results is that
only a subset of Runx2 target genes is important for osteoblast biology, and it shares
no overlap with the subset of known Runx2 target genes. Thus, an important step in
unraveling the mechanism of Runx2 will be to identify novel target genes in
osteoblasts.

Although identifying novel Runx2 target genes would greatly improve our
understanding of its mechanism in osteoblasts, such an understanding would not be
complete without a better grasp of how Runx2 regulates its target genes. In general,
it is appreciated that Runx2 is recruited to genomic sites, where it can influence the
transcription of nearby genes via interactions with a myriad of co-factors, including
p300 [Sierra et al., 2003], MORF and MOZ [Pelletier et al., 2002], Histone
deacetylase 6 [Westendorf et al., 2002], histone deacetylase 3 [Schroeder et al.,
2004] HDACs 4 and 5 [Kang et al., 2005], Grg5 [Wang et al., 2004], retinoblastoma
protein [Thomas et al., 2001], HES1, [Lee et al., 2006], p204 [Luan et al., 2007],
interferon-inducible p204 [Liu et al., 2005], Smads 3 & 2 [Alliston et al., 2001;
Selvamurugan et al., 2004], Smad 1 and Smad 5 [Hanai et al., 1999; Lee et al., 2000;
Nishimura et al., 2002], Lef-1 and Tcf-4 [Kahler and Westendorf, 2003; Reinhold
and Naski, 2007], Msx2 and Dlx5 [Roca et al., 2005; Shirakabe et al., 2001], Msx2

interacting nuclear target protein [Sierra et al., 2004], 1 ,25-dihydroxyvitamin D
3

receptor, [Paredes et al., 2004], Hes-1 [Shen and Christakos, 2005], the estrogen
8
receptor [McCarthy et al., 2003], ETS1 [Sato et al., 1998], ATF4 [Xiao et al., 2005],
and YAP [Zaidi et al., 2004]. The direction and magnitude of the target gene’s
response depends likely depends on the particular co-factor(s) involved. Meanwhile,
the complement of co-factors that interact with Runx2 is likely dependent on cell
type, developmental stage, and promoter context.

In contrast to the plethora of literature surrounding interactions between Runx2 and
co-regulatory molecules, relatively little information exists regarding its recruitment
to genomic sites. In general, it is appreciated that this occurs via interaction with
sequences resembling 5’-TGTGGT-3’, or its complement 5’-ACCACA-3’, although
these sequences are found at such high frequency throughout the genome that they
are probably not sufficient for binding [Roca et al., 2005]. Moreover, the binding
sequence may not even be necessary in some cases as Runx2 can be recruited to
genomic sites via interaction with co-factors [Cui et al., 2003; Gutierrez et al., 2002;
Shimoyama et al., 2007; Shin et al., 2006; Ziros et al., 2004]. Beyond the mystery
surrounding the exact signals that target Runx2 to its appropriate genomic sites, the
dynamics between recruitment and target gene expression are unclear, as are the
dynamics between recruitment and expression of Runx2 itself. There may be simple
positive correlations between these parameters, but this is not necessarily a given.
For example, Runx2 genomic recruitment may be disconnected from its expression
via regulation at the post-transcriptional or post-translational levels. Moreover,
binding to genomic sequences may not correlate well with changes in nearby gene
9
expression, due to absence of co-regulatory factors. Finally, these dynamics may be
context-dependent, with different trends occuring at different targets.

The following paper describes our work towards unraveling the mechanisms of
Runx2 action in osteoblasts. It begins with the identification of several novel Runx2
genomic targets by ChIP Display, along with several putative Runx2 target genes.
One of these genes, Tram2, is selected for further characterization based on its
postulated role in type I collagen synthesis. We demonstrate that its expression
pattern is consistent with the hypothesis that it is regulated by Runx2. This paper
then goes on to describe our characterization of the dynamics between expression of
Runx2 and a known target gene during osteoblast development. We show that there
is a surprising incongruence between these two parameters, and that stimulation of
target gene expression is not attributable to changes in Runx2 expression. Rather, it
may be attributed to genomic recruitment of Runx2, which increases during
development of the osteoblast phenotype. Finally, this paper describes the
identification of another novel Runx2 genomic target, this time by ChIP-Chip. We
show that Runx2 occupancy at this locus increases during osteoblast development as
well, suggesting that a common regulatory mechanism governs the recruitment of
Runx2 to its genomic targets.
10
Chapter 2: Materials and Methods

Cell Culture
A subclone derived from the MC3T3-E1 osteoblastic cell line

was used in this study
[Smith et al., 2000]. Cells were maintained

in –MEM supplemented with 10% fetal
bovine serum (FBS)

(Invitrogen, Carlsbad, CA). Starting at confluence, 10 mM
sodium ß-glycerophosphate

and 50 µg/ml ascorbic acid (Sigma, St. Louis, MO) were
added to support

differentiation. ST2 cells (Riken, Tsukuba, Japan) were maintained
in RPMI1640

supplemented with 10% FBS, while C3H10T1/2 and C2C12 cells were
maintained in DMEM supplemented with 10% FBS. Primary skeletal myoblasts
were cultured as previously described [Gersbach et al., 2006]. When indicated, 300
ng/ml rhBMP-2 was added starting at confluence and at each subsequent feeding.

Runx2 over-expression
Runx2 was ectopically expressed in primary myoblasts by viral infection as
previously described [Gersbach et al., 2006]. C3H10T1/2 cells were transiently
transfected using Lipofectamine 2000 (Invitrogen) with the Runx2 vector pCMV-Osf
[Ducy et al., 1997], or the molar equivalent of pcDNA3 (Invitrogen). pCAT
(Promega, Madison, WI) was used to equalize the total amount of DNA that was
transfected.

11
Statistical Analysis
Results from each quantitative assay

were analyzed by Student’s

t-Test to ascertain
the effect of Runx2 over-expression and/or rhBMP-2 treatment on gene expression.
Differences were considered significant when

p ≤ 0.05.

RNA Analysis
RNA was collected from cells using the Aurum Total RNA Mini Kit (Bio-Rad,
Hercules, CA) and quantitated by NanoDrop (Thermo Scientific, Waltham, MA).
One microgram was then used to generate cDNA with the SuperScript III cDNA
synthesis kit (Invitrogen). Two microliters from a 1:15 dilution of the resulting
cDNA was used for real-time PCR with iQ
TM
SYBR Green Supermix (Bio-Rad).
Real-time PCR was performed on an iCycler with MyiQ single-color detection
system (Bio-Rad). Primers sequences are given in Tables I & III.

Protein Analysis
Cell extracts were prepared using the Nuclear Extraction Kit (Active Motif,
Carlsbad, CA) following the Preparation of Whole-Cell Extract protocol. Protein
concentration was determined with the Bio-Rad Protein Assay Kit (Bio-Rad) and 10
µg were used for Western blot analysis. Gel shift and super-shift assays, were
performed essentially as described with 20 µg protein [Luppen et al., 2003]. Anti-
Runx2 antibodies #10758 (Santa Cruz, CA) were used for both Western and super-
shift analyses.
12

ChIP assays
ChIP was performed essentially as described [Jia et al., 2003]; however, for the
MC3T3-E1 cells, we found that an extend period of sonication (approximately 6
min) was required to achieve the optimal average chromatin fragment size of 500-bp.
Primers used to assess Runx2 occupancy at specific genomic sites are listed in Table
I & II.

ChIP-chip
ChIP samples were amplified using the Whole Genome Amplification Kit (Sigma),
labeled, and hybridized to NimbleGen array #7 of the MM8 tiling set (NimbleGen,
Madison, WI). Peak calling was performed as described [Jia et al., 2008]. Despite
excellent technical reproducibility, only a few peaks were validated by qPCR or
shared by an independent biological replicate, demonstrating possible limitations of
Runx2 ChIP-Chip analysis.

ChIP Display
A detailed description of ChIP Display is provided in the following chapter, and has
been published elsewhere [Barski et al., 2008]

13

Table I. Oligonucleotides used in Chapter 4
Name Sequence
Short CD Linker 5’ –TTC GCG GCC GCA C- 3’
Long CD A Linker 5’ –GAC GTG CGG CCG CGA A- 3’
Long CD T Linker 5’ –GTC GTG CGG CCG CGA A- 3’
“AA” CD primer 5’ –CGG CCG CAC GAC CA- 3’
“AT” CD primer 5’ –CGG CCG CAC GAC CT- 3’
“AG” CD primer 5’ –CGG CCG CAC GAC CG- 3’
“AC” CD primer 5’ –CGG CCG CAC GAC CC- 3’
“TA” CD primer 5’ –CGG CCG CAC GTC CA- 3’
“TT” CD primer 5’ –CGG CCG CAC GTC CT- 3’
“TG” CD primer 5’ –CGG CCG CAC GTC CG- 3’
“TC” CD primer 5’ –CGG CCG CAC GTC CC- 3’
Tram2 forward ChIP primer 5’ –AGC TCT GCA ATT GGT TCG- 3’
Tram2 reverse ChIP primer 5’ –TGT CCC GCA CGT TAT CTG- 3’
Insulin forward ChIP primer 5’ –AAC TGG TTC ATC AGG CCA TCT GGT C- 3’
Insulin reverse ChIP primer 5’ –TGG ATG CCC ACC AGC TTT ATA GTC C- 3’
Osteocalcin forward ChIP primer 5’ –GAG AGC ACA CAG TAG GAG TGG TGG AG- 3’
Osteocalcin reverse ChIP primer 5’ –TCC AGC ATC CAG TAG CAT TTA TAT CG- 3’
L10A forward qRT-PCR primer 5’ –CGC CGC AAG TTT CTG GAG AC- 3’
L10A reverse qRT-PCR primer 5’ –CTT GCC AGC CTT GTT TAG GC- 3’
Runx2 forward qRT-PCR primer 5’ –AGC CTC TTC AGC GCA GTG AC- 3’
Runx2 reverse qRT-PCR primer 5’ –CTG GTG CTC GGA TCC CAA- 3’
OC forward qRT-PCR primer 5’ –CGG CCC TGA GTC TGA CAA A- 3’
OC reverse qRT-PCR primer 5’ –GCC GGA GTC TGT TCA CTA CCT T- 3’
Tram2 forward qRT-PCR primer 5’ –CCC CGA GAA AGG GAA CTT TA- 3’
Tram2 reverse qRT-PCR primer 5’ –TTC TCT GCC TTC ACC ACT CC- 3’

Table II. Oligonucleotides used in Chapter 7
A. EMSA Probes
OSE2 5’-AGCTGCAATCACCAACCACAGCA-3’ 5’-TGCTGTGGTTGGTGATTGCAGCT-3’
B. RT-qPCR Primers
OC 5’-TGCCAGAGTTTGGCTTTAGG-3’ 5’-CATGAGGACCCTCTCTCTGC-3’
Runx2 5’-GCCTTCAAGGTTGTAGCCC-3’ 5’-CCCGGCCATGACGGTA-3’
rpL10A 5’-CGCCGCAAGTTTCTGGAGAC-3’ 5’-CTTGCCAGCCTTGTTTAGGC-3’
C. ChIP Primers
OC 5’-CTAATTGGGGGTCATGTGCT-3’ 5’-CCAGCTGAGGCTGAGAGAGA-3’
Insulin 5’-TAGCACCAGGCAAGTGTTTG-3’ 5’-CTGCTTGCTGATGGTCTCTG-3’
Glt28d2 5’-ATGCTGTGGTGGAAAACCTC-3’ 5’-AGCTCTCCCCTTCAGTCTCC-3’
14
Chapter 3: Identification of Transcription Factor Target Genes by
ChIP Display (CD)

3.1 Introduction
Transcription factors play pivotal roles in cell biology, animal physiology and
disease processes. The ongoing quest to understand their function includes the
search for genes whose expression they regulate. Target genes for transcription
factors can be identified by expression studies, although these are fraught with
problems (see Note 1). A complementary group of approaches is based on the
physical interaction of transcription factors with cis-acting regulatory elements, and
the assumption that such elements are frequently located close to the genes that they
regulate. Most of these approaches begin with chromatin immunoprecipitation
(ChIP), in which DNA fragments bound by a transcription factor of interest in living
cells are immunoprecipitated with antibodies against that protein. Once identified,
these fragments can be mapped to the genome, and nearby genes can be tested for
their regulation by the transcription factor of interest.

The ChIP-based methods for transcription factor target discovery are challenged by
the overwhelming excess of non-specifically precipitated fragments (Fig. 1A). Each
of these methods takes a different path to identifying those fragments that were
specifically immunoprecipitated. Of these methods, ChIP Display (CD) [Barski and
Frenkel, 2004] is one of the least high-throughput but offers relative simplicity and
15
can be performed in a typical molecular biology laboratory without sophisticated
equipment or complicated statistical analyses (see Note 2). To overcome the high
background problem, CD effectively concentrates fragments representing each target
while scattering the remaining DNA. Targets are concentrated via restriction
digestion (Fig. 1C): all fragments representing a certain target now have the same
size, allowing one to resolve them as a single band on a gel. Scattering of the non-
specifically precipitated fragments is achieved by dividing the total pool of
restriction fragments into families based on the identity of nucleotides at the ends of
such fragments (Fig. 1E-G). Because all restriction fragments representing each
given target have the same nucleotides at the ends, they remain in the same family
and the signal is not eroded. In contrast, the other fragments, mostly background, are
scattered into many families.

The CD protocol is rather simple (Fig. 1). Following ChIP, immunoprecipitated
DNA is dephosphorylated by shrimp alkaline phosphatase (SAP) to prevent ligation
of linkers to DNA ends generated by sonication (Fig.1B). SAP is then heat-
inactivated and the DNA is digested with a restriction enzyme. We use AvaII, whose
recognition sequence, GGWCC (W=A or T), can be expected approximately every
500 bp in a random sequence. This is followed by ligation-mediated PCR (LM-
PCR) using various primer combinations (Fig. 1E). Up to thirty-six combinations of
eight nested primers are employed to amplify fragments belonging to one family at a
time (Fig. 1F, G). Each such primer contains either A or T at the +3 position of the
16
AvaII site and one nested nucleotide, A, T, G or C, at the 3’end (Fig. 1F). Amplified
fragments from two to three independent immunoprecipitates and two to three
control precipitates are resolved by PAGE. Bands enriched in the ChIP lanes are
considered candidate targets and are excised from the gel for further characterization.
Using CD, our lab has identified dozens of targets for both Runx2 and the androgen
receptor [Barski and Frenkel, 2004; Jariwala et al., 2007; Pregizer et al., 2007;
Prescott et al., 2007].
17
Fig. 1. Principles of ChIP Display (CD). See next page (A) Precipitated DNA
fragments (green – specific; black – nonspecific) are aligned with the genome.
Graph (red) describes the representation of each nucleotide in the immunoprecipitate.
The 1-Mb area shown contains two hypothetical target genes, #1 and #2. (B)
Magnification of the two regions of interest from A. DNA fragments are treated
with shrimp alkaline phosphatase (SAP) to prevent linker ligation to DNA ends
generated during sonication. (C) DNA is digested with AvaII. (D) Linkers are
ligated to the ends of AvaII fragments. Also shown are the nested primers. Note that
different primers will amplify targets #1 and #2 (see color plate). (E) PCR products
amplified in three reactions: Left – Target #1 is amplified in a reaction with a single
primer (family X in panel G). Middle – Target #2 is amplified with two different
primers (family V in panel G); note that target #1 is amplified here again; Right –
Most of the PCR reactions will amplify neither target #1 nor target #2; however, they
will amplify targets from other loci. (F) Linkers and nested primers. Positions +1 to
+6 are defined at the top. Nucleotides at positions +3 and +6 (highlighted) are used
to segregate AvaII fragments into families. Linkers (blue) contain a 1:1 A:T mixture
(W) at position +3 and a C at position +1 to destroy the AvaII sites. In addition, this
panel shows one of the eight PCR nested primers, the one with T at position +3 and
C at position +6, which would amplify the family indicated by X in panel G. (G)
Thirty-six families of fragments, each amplified using one or two of eight nested
primers. White squares correspond to a single primer and shaded squares correspond
to two different primers.
18
B A
N GG
T
C C
AvaII
+1+2
+3
+4+5+6
A
T
G
C
A
T
G
C
A
A
A
A
T
T
T
T
+6
+6 +3
GGWCG
C
C
C
T
T
C GWCC
GCWGG
+3 +6
X V
A T G C A T G C
A A A A T T T T
G F
A
A
T
G
C
A
T
G
C
A
A
A
A
T
T
T
T
G
CCWGC G
X V
A T G C A T G C
A A A A T T T T
+3
A
B
1Mb
C
D
5kb
5kb
E
5kb
5kb
5kb
5kb
PCR
1Mb 1Mb
1Mb
G
#1
#2
SAP
AvaII
Ligate
linkers
A
B
1Mb
C C
D D
5kb
5kb
E E
5kb
5kb
5kb
5kb
PCR
1Mb 1Mb
1Mb
#1
SAP
AvaII
Ligate
linkers
#2
19
3.2 Materials
Digestion and Linker Ligation
1. Short linker oligo: 5’-TTCGCGGCCGCAC-3’
2. Long A linker oligo: 5’-GACGTGCGGCCGCGAA-3’
3. Long T linker oligo: 5’-GTCGTGCGGCCGCGAA-3’
4. 10X React1 and 10X React2 buffers (Invitrogen)
5. Shrimp Alkaline Phosphatase (SAP), 1 unit/µl
6. AvaII, 10 units/µl
7. T4 DNA ligase, 400 units/µl
8. 10X T4 Ligase Buffer
9. Enzymatic Reaction Cleanup Kit (Qiagen)

PCR Amplification
1. Taq DNA polymerase, 5 units/µl
2. 10X PCR Reaction Buffer, -MgCl
2

3. MgCl
2
, 50 mM
4. dNTP mix (Invitrogen), 10 mM each
5. Up to eight PCR primers: 5’-CGGCCGCACGWCCN-3’, where W is A or T
and N is A, G, C, or T (see Note 2).

PAGE
1. 40% Polyacrylamide solution [29 :1 (w/w) acrylamide/bisacrylamide]
20
2. 10% (w/v) ammonium persulfate (APS) in water (store < 1 month at 4°C)
3. TEMED solution (N,N,N,N-tetramethylethylenediamine; store at 4°C)
4. 10X TBE Electrophoresis Buffer, pH 8.0 (890 mM Tris, 890 mM boric acid
and 20 mM EDTA)
5. 5X Sample Loading Buffer (10% Ficoll 400, 50 mM disodium EDTA (pH
8.0), 0.5% Sodium dodecyl sulfate, 0.25% bromphenol blue, 0.25% xylene
cyanol)
6. Gel loading pipet tips
7. 1-kb DNA Ladder

Target Identification
1. Spin-X Centrifuge Tube Filters (Corning)
2. HaeIII, HinfI, MspI, RsaI, 10 units/µl each
3. Standard/High Melt Agarose
4. Gel extraction kit (Qiagen)
5. BigDye
®
Terminator 3.1 Cycle Sequencing Kit (Applied Biosystems)
6. AutoSeq
®
G-50 Microcentrifuge Columns (Amersham)
7. Hi-Di™ Formamide (Applied Biosystems)
8. For in-house sequencing: ABI PRISM® 3100 Genetic Analyzer (Applied
Biosystems) or an alternative system.

21
3.3 Methods
CD begins after ChIP has been completed with the cells and antibodies of interest.
To identify Runx2 targets in osteoblasts, we usually ChIP half the chromatin
collected from two confluent 100 mm plates of MC3T3-E1 cells with 5 µg of Runx2-
specific antibodies (Santa Cruz, Cat No. 10758X). The other half is
immunoprecipitated with 5 µg of non-specific IgG to control for non-specific IP.
Each of the specific and non-specific ChIPs is done in duplicate or triplicate, to
control for experimental variation. Thus, we typically initiate a CD experiment for a
given experimental condition with two or three ChIP samples and two or three mock-
ChIP samples. Each of the samples is cleaned with an Enzymatic Reaction Cleanup
Kit (Qiagen) and eluted in 50 µl H
2
0, which is sufficient for about 100 separate CD
reactions. Each CD reaction is then displayed in one lane of an acrylamide gel, and
will typically yield up to two targets.

Although we do not describe the ChIP procedure in this chapter, two critical points
must be emphasized: (i) No salmon sperm DNA should be used for ChIP intended
for CD. Instead, substitute with bacterial tRNA (see Note 3); (ii) High-quality ChIPs
are essential for the success of CD. Before a CD protocol is initiated, one must
ensure that the starting ChIP material is enriched for at least one, and preferably
more, known targets. We recommend the use of real-time PCR to determine the
enrichment levels, with both internal (non-targets) and external (IgG) controls (see
22
Note 4). If the enrichment levels are satisfactory (~10-fold or better), then this
material may be used for CD.

Digestion and Linker Ligation
1. Resuspend each of the two long linker oligos (“A” and “T”) and the short
linker oligo in dH
2
0 at a 100 µM concentration.
2. Mix 15 µl of the short linker oligo with 15 µl of the long “A” linker oligo.
Add 10 µl of 10X React2 buffer and 60 µl of dH
2
0. Likewise, mix 15 µl of
the short linker oligo with 15 µl of the long “T” linker oligo and add 10 µl of
10X React2 buffer and 60 µl of dH
2
0. Mix each well.
3. Anneal oligos by placing samples in a beaker of boiling water for 1 minute.
Turn off heat and allow water to return to room temperature without
removing samples. After cooling, annealed linkers may be stored at -20°C.
4. Mix 10 µl of each ChIP sample with 2 µl of 10X React1 buffer, 1 µl of SAP,
and 7 µl of H
2
O. Incubate at 37°C for 30 minutes. Heat denature for 15
minutes at 65°C (see Note 5).
5. Add 1 µl (10 units) of AvaII to each sample. Incubate at 37°C for 30 minutes
(see Note 6).
6. Add 2 µl each of both annealed linkers, 3 µl of 10X T4 ligase buffer and 1 µl
T4 DNA ligase (400 units). Incubate overnight at 16°C.
23
7. Add 0.5 µl of AvaII and 0.5 µl of T4 Ligase to each sample. Incubate 1 hour
at room temperature (see Note 7). Purify the samples on QuickSpin columns
using Enzymatic Reaction Clean-Up Buffer. Elute with 50 µl dH
2
0.

PCR Amplification
1. Make a PCR master mix containing 1X PCR Buffer, 1.5 mM MgCl
2
, 0.2 mM
each dNTP, 0.025 units/l Taq DNA Polymerase (see Note 8), and 0.5 µM of
one or two of the eight possible primers (see Note 9).
2. Add 18 µl of the PCR master mix to 2 µl of each replicate of each column-
purified ChIP sample (see Notes 10 and 11).
3. Mix by pipetting up and down, then place tubes in a thermocycler at room
temperature (see Note 12).
4. Amplify the three samples representing each IP using two or three annealing
temperatures between 68°C and 70°C. Specifically, amplify each set of three
samples using the following program: 1) 72°C for 30 sec; 2) 95°C for 5 min;
3) 95°C for 1 min; 4) 68°C [or 69°C or 70°C] for 1 min; 5) 72°C for 1 min;
6) Go to 3, 45 times; 7) 72°C for 5 min; 8) End (see Figs. 2A and 3, and Note
13).

PAGE and Target Isolation
We recommend that the CD procedure be completed with one or two of the 36
primer combinations (represented by the 36 boxes in Fig. 1G) before moving on to
24
additional families. Only very experienced investigators should consider performing
all of the LM-PCR reactions with all possible primer combinations (Fig. 1G) prior to
PAGE and target identification.
1. Cast a 1X TBE, 8% native polyacrylamide gel for resolving nucleic acids.
We use gels 14 cm long and 17 cm wide.
2. After polymerization, load the wells with 1X sample loading buffer and pre-
run in 1X TBE for 30 minutes at 150 volts.
3. Add 5 µl of 5X sample loading buffer to each sample after amplification and
mix well.
4. Flush the wells of the gel thoroughly with a syringe. Using gel loading
pipette tips, carefully apply 10 µl of each sample to the bottom of each well.
We typically load samples immunoprecipitated with specific antibodies on
one half of the gel, while samples immunoprecipitated with non-specific
antibody are loaded on the other half (Fig. 2A). We load replicates next to
each other and place molecular weight markers in the middle and on each
side of the gel.
5. Allow the gel to run at 150-200 volts, until the xylene cyanol dye is 3 cm
from the bottom of the gel.
6. Remove the gel and incubate in 1X TBE supplemented with 100 ng/ml EtBr.
Visualize bands using a UV transilluminator. Capture and print the image.
25
7. Note any co-migrating bands that are reproducible in the ChIPs with the
specific antibody. If these do not co-migrate with bands in the control lanes,
then they are considered candidate targets (see Figs. 2A, 3, and Note 14).
8. Cut the bands that represent good candidate targets out of the gel using a
fresh blade for each band. Be careful to avoid neighboring bands and lanes.
9. Place each excised gel piece in a Spin-X column, and freeze at -80°C. Spin
the columns at maximum speed in a microcentrifuge for 15 minutes at room
temperature.

26
Fig. 2. Target Display and Identification. (A) PAGE of Runx2 targets in one of the 36 CD families shown in
Fig. 1G. DNA samples from three independent ChIPs and three mock ChIPs were subjected to LM-PCR with
one of the 36 possible primer combinations (Fig. 1G). Each PCR was performed with an annealing temperature
of either 68°C or 69°C as indicated. The products were then resolved side-by-side on an 8% polyacrylamide gel.
Bands indicated with white arrowheads are not putative targets because they are comparably present in both the
ChIP and the mock lanes. Bands representing a putative target are indicated with black arrowheads. The bands
labeled I and ii were excised from the gel, purified and re-amplified. (B) The re-amplified products from panel A
were treated with the restriction enzymes HaeIII (Ha), HinfI (Hi), and MspI (Ms). The digestion products were
resolved on a 4% agarose gel. This example demonstrates that targets can be identified even in the presence of
contaminating bands, seen in the lanes containing undigested products (U). The bands of interest, identified
based on their size (compare to panel A), are indicated with arrows. A sub-fragment likely to have arisen from
the band of interest is indicated in each of the HinfI lanes with an arrowhead. Indeed, after sequencing these
AvaII-HinfI sub-fragments and after mapping of the “hit” to the mouse genome, it was shown to reside between
two AvaII sites separated by 473 bp, consistent with the size of the band of interest that was excised from the
acrylamide gel (Panel A, and see section 3.4.11). M, Molecular weight marker.

Target Identification
In principle, targets can be identified by cloning into a vector followed by
sequencing. However, we prefer to identify targets without cloning, by directly
sequencing sub-fragments obtained through secondary digestion as described below
(see Note 15).
1. Re-amplify the eluted DNA using the same primers and conditions as in
Section 3.2, in a total reaction volume of 50 µl.
2. Split each amplified DNA sample into five 10 µl aliquots. Digest 2-4 of
these aliquots with different restriction enzymes, each in a final volume of 20
µl (see Note 16).
M 69° 68°
1
ChIP mock
69° 68° M 69° 68° 69° 68° 69° 68° M 69° 68°
2 3 1 2 3
A.
i. ii.
517
506
396
344
298
bp
M U Ha Hi Ms M U Ha Hi Ms
i. ii. B.
517, 506
396
344
298
bp
220
201
154
134
27
3. After digestion, add 5 µl of 5X sample loading buffer to each sample, and
mix thoroughly. Load 10 µl of each sample on a 4% agarose gel containing
100 ng/ml ethidium bromide. Load the undigested sample in lane #1,
followed by the digested samples (see example in Fig. 2B). Repeat this for
each group of samples, separating the groups with 1-kb DNA ladder. Allow
the gel to run at 100 volts for 2-3 hours.
4. Place the agarose gel on a UV transilluminator, and note the pattern of bands
(sub-fragments) resulting from the restriction digestion for each candidate
target. Look for identical patterns of sub-fragments, which appear to have
originated from a common target. Excise several such sub-fragments from
the gel for sequencing (see Note 16).
5. Purify the DNA on MinElute columns (Qiagen) according to the
manufacturer’s protocol. Sequencing of the eluted DNA can be performed
either in house or using a service lab. Steps 6-9 below describe the
sequencing procedure used in our lab.
6. Mix 10 µl of each eluted DNA sample with 2 µl 5X BigDye Sequencing
Buffer, 4 µl Ready Reaction Premix, and 5 pmol of the primers used in Step
3.2.1. Add dH
2
0 up to 20 µl. Place samples in a thermocycler and run the
following program: 1) 95°C for 4 min; 2) 93°C for 30 sec; 3) 50°C for 15
sec; 4) 60°C for 4 min; 5) Go to 2, 30 times; 6) End.
7. Remove samples from the thermocycler and clean DNA with AutoSeq
®
G-50
Microcentrifuge Columns.
28
8. Place samples in SpeedVac and dry under low heat. Resuspend in 10 µl Hi-
Di™ Formamide.
9. Transfer samples to a 96-well plate. Heat to 100°C for 5 minutes and cool
immediately on ice for 2 minutes. Load plate into ABI PRISM® 3100
Genetic Analyzer and begin run.
10. Map the resulting sequences within the genome of interest, using BLAST or a
similar search engine.
11. Search the regions surrounding the “hit” for AvaII recognition sequences.
The distance (in bps) between the two sites on either side of the sequenced
region should closely match the size of the fragment that was excised from
the polyacrylamide gel. Additionally, the identities of the nucleotides at the
ends of each fragment should match the primer pair used to generate the
fragment in Section 3.2. Finally, the fragment should contain restriction
enzyme recognition sequences at positions that are compatible with the
digestion patterns that were seen on the agarose gel (step 3.4.4).
12. Validate each “hit” of interest using conventional ChIP assay with locus-
specific primers (see Note 17). Annotated genes near confirmed targets are
potentially regulated by the transcription factor of interest.

3.4 Notes
1. Target genes can be identified in comprehensive gene expression studies
(e.g., using microarrays) based on their response to over-expression or
29
activation (e.g., by ligand) of transcription factors. However, there are at
least four problems with expression-based approaches. First, it is generally
difficult to tell whether a responsive gene is a direct or an indirect target of
the transcription factor of interest. Second, in experiments involving over-
expression of transcription factors, the response of some genes may be forced
by exaggerated concentrations of the protein, resulting in physiologically
insignificant results. Third, such studies do not provide information on the
location of the cis-acting regulatory elements. Last, expression studies are
unable to disclose genes, to which the transcription factor of interest binds
without functional consequences under the experimental conditions
employed. Such genes may become responsive to the transcription factor
under different conditions, for example in the presence of a specific
extracellular stimulus and/or a particular co-activator (either in the same or in
a different cell type).
2. CD is a “sampling” method and will only disclose a fraction of all the targets.
High throughput methods for transcription factor target discovery include
ChIP-Chip [Cawley et al., 2004; Horak and Snyder, 2002; Ren et al., 2000],
SABE [Chen and Sadowski, 2005], STAGE [Kim et al., 2005], ChIP-PET
[Wei et al., 2006], GMAT [Roh et al., 2004], SACO [Impey et al., 2004], and
DamID [van Steensel and Henikoff, 2000]. These methods are also used to
map histone modifications, a task for which CD is not well suited. However,
the high throughput methods share requirements, which render them
30
prohibitively expensive for many labs. First, the amount of starting material
needed is relatively large. Second, complex statistical analysis is needed to
discern true targets from the large number of non-specifically
immunoprecipitated fragments. Third, complicated equipment is required
and the price is extremely high. This is exacerbated by the possibility of poor
reproducibility, which usually necessitates many repetitions, especially in the
beginning of a project with a new transcription factor. In addition, unlike
some of the high throughput methods listed above, CD can lead to novel
targets before a complete screen is finished. Of the 36 reaction types that
constitute a full screen (Fig. 1G), one can start with one or a subset of
reactions and continue only if more targets are desirable. Furthermore, CD
provides the opportunity to compare occupancy between different cells, and
to pursue only those with interesting phenotypes. For example, one can
display side-by-side targets occupied in two different cell types, or under
different physiological conditions, and then pursue the identification of only
selected targets.
3. Any ChIP protocol can be adapted for CD. However, the typical high
background of non-specifically immunoprecipitated fragments, which can be
tolerated in conventional ChIP assays of candidate targets, is the Achilles’
heel of methods intended to identify unknown targets. Although CD is
relatively forgiving in this sense, it will not distinguish between targets and
contaminating DNA. Therefore, extra care should be taken to prevent
31
contamination. Related to this, the beads used for immunoprecipitation
should be preadsorbed with bacterial tRNA instead of salmon sperm DNA.
tRNA will not be ligated or amplified during the following steps.
Alternatively, the preadsorption step can be skipped altogether.
4. Enrichment for known targets is a measure of the success of a ChIP
experiment. Using real-time PCR with locus-specific primers, one compares
the concentrations of known targets to those of non-targets after ChIP. To
more rigorously measure the enrichment factor, the known-target-to-non-
target ratio in the ChIP is compared to the respective ratio in samples
immunoprecipitated with non-specific antibodies, which should in theory be
equal to 1 for single-copy genes. In our lab, we achieve a ~10-fold
enrichment for Runx2 targets in MC3T3-E1 cells.
5. Prior to AvaII treatment, we dephosphorylate the DNA with SAP to prevent
ligation of linkers to the ends produced by sonication. We include this step
even though we are not sure that sonication produces ends that are
compatible with ligation. SAP was chosen because it can be efficiently heat
denatured to prevent dephosphorylation of fragments generated later for LM-
PCR by AvaII digestion.
6. A key feature of CD is the restriction digestion, a step that standardizes the
size of a diverse group of DNA fragments obtained from a given target after
sonication and ChIP. The ability of CD to disclose a given target depends on
the presence of two recognition sites for the restriction enzyme in the vicinity
32
of that target. We use the 5-bp cutter AvaII because the genomic distribution
of its cognate sites can be expected to result in many sonicaiton fragments
(which themselves are on the order of 300-900 bp) containing two AvaII
sites. Sites for 6-bp cutters occur too infrequently, so it is unlikely that two
sites would be present in one sonication fragment. Of course even for 5-
cutters such as AvaII, many sonication fragments will not be long enough and
targets will be missed. Digestion with 4-bp cutters, however, would result in
a large number of fragments and would unnecessarily increase the
complexity of the DNA fragment pool. Of several possible 5-bp cutters, we
chose AvaII because it is active in salt-free React1 buffer, eliminating the
need for desalting prior to ligation. Conveniently, AvaII is an inexpensive
and stable enzyme. Noteworthy, digestion by AvaII can be blocked by CpG
methylation. While this enzyme property may limit target discovery, it can
be considered advantageous because methylated regions may be associated
with less accessible genes. Be that as it may, among the initial Runx2 targets
identified by CD, a fair number were associated with CpG islands.
7. The second dose of AvaII is added to digest concatemers of AvaII fragments
that might have been generated during linker ligation. Re-digestion with
AvaII recycles these concatemers, while the presence of ligase facilitates
linker ligation. Because the linkers do not restore the AvaII site (Fig. 1F),
their ligation to digested DNA fragments is irreversible even in the presence
of AvaII.
33
8. We use Taq DNA Polymerase because it lacks proofreading activity. This
prevents “editing” of the nested primers, which may anneal to the template
even when the nucleotides at positions +3 or +6 (Fig. 1F) are not
complementary. Such mismatches would be perfect substrates for a 3’-5’
proofreader, which would then use the edited primer to amplify the fragment
even when it should not be amplified in that specific reaction.
9. By amplifying targets in individual families (Fig. 1G), we increase the signal-
to-noise ration for one group of targets at a time. Because all fragments
originating from a given target have the same nucleotides at the ends, they
remain in the same family and the signal is not eroded. In contrast, the
background fragments are scattered into many families (Fig. 1E). This
feature of CD also provides investigators with the choice of performing either
a full or a partial screen, i.e., using as many primer combinations (families) as
they wish. In our search for Runx2 target genes in MC3T3-E1 cells, we
limited the initial CD screen to eight families, those that are amplified by a
single primer each [Barski and Frenkel, 2004], and later continued with other
families [Pregizer et al., 2007].
10. In addition to biological replicates, we strongly encourage the use of
technical replicates. There is a huge number of non-specifically
immunoprecipitated fragments in each ChIP sample that are all competing for
a limited number of primers in the amplification reaction. Usually, a subset
of these fragments wins out early on in the competition, often for random
34
reasons. The so-called “founder effect” and the complexity of the template
DNA result in a different pattern of amplification from run-to-run of the same
sample (Fig. 2A). Although fragments that represent true targets are present
at a higher concentration and should be amplified reproducibly, technical
replicates help one to distinguish between true targets and spuriously
amplified fragments (Fig. 2A).
11. As an alternative to using direct replicates, it may be worthwhile to perform
the PCR on different dilutions of the ChIP material. Seeing a particular band
with all dilutions enhances one’s confidence that it truly represents a target
(Fig. 3A). Other bands, which are visible with lower dilutions, but not with
higher dilutions (see example in Fig. 3B) can be either dismissed or pursued
contingent upon validation (step 3.4.12).
12. Do not pre-heat the thermocycler, because this might compromise
amplification efficiency. The linkers used for CD are unphosphorylated
oligonucleotide duplexes, thus only the short oligo of the duplex is ligated to
the AvaII fragment. The long oligo is kept in place through non-covalent
interactions and will be lost upon heating. To allow the polymerase to fill-in
the lost sequence before the strands are completely separated, we start at
room temperature and hold the reaction for 30 seconds at 72°C. Hot-start
polymerases complexed with antibodies should not be used in this PCR as
they become active only after the reaction is heated to 95°C.

35
M - + - + - + - + - + - + - + - + - + M
68°C 69°C 70°C 68°C 69°C 70°C 68°C 69°C 70°C
1:1 1:4 1:16
A.
517
506
396
344
298
bp
M - + - + - + - + - + - + - + - + - + M
68°C 69°C 70°C 68°C 69°C 70°C 68°C 69°C 70°C
1:1 1:4 1:16
B.
517
506
396
344
298
bp
Fig. 3. Effect of Dilution and Annealing Temperature on CD. (A) Three dilutions (1:1, 1:4, 1:16)
of Runx2 ChIP (+) or mock ChIP (-) were amplified with one of the 36 possible primer combinations
(Fig. 1G) at an annealing temperature of 68°C, 69°C, or 70°C, and the products were resolved on an
8% non-denaturing polyacrylamide gel. Arrowheads indicate a putative target that was reproducible
at all dilutions, and with both 70°C and 69°C, but not 68°C, as the annealing temperature. (B)
Another example, using a different primer set. Arrowheads indicate a putative target that appeared
only with annealing temperatures of 69°C and 70°C, and only without dilution. Although this
candidate is inferior to the one in Panel A, it can be considered a putative target, but might very well
prove to be a false positive during the validation steps (sections 3.4.11 and 3.4.12). M, Molecular
weight marker.

13. Due to the large amount of non-specifically precipitated fragments in a ChIP
sample, non-specific amplification is a technical challenge that must be
overcome by CD. To this end, we use short (14-bp) primers and high
36
annealing temperatures. Unfortunately, it is possible for the annealing
temperature selected to be too stringent for any amplification to occur.
Because it is difficult to predict what annealing temperature will yield the
right balance between too stringent and too permissive, we recommend that
replicates be performed at slightly different annealing temperatures. This can
be accomplished easily using a thermocycler with gradient capabilities. Two
or three temperatures between 68°C and 70°C usually provide good results
(See Figs. 2, 3).
14. In theory, targets should appear as co-migrating bands in all of the ChIP lanes
but none of the mock ChIP lanes. In practice, such perfect scenarios are
infrequent (see Fig. 2A), and investigators must exercise their own good
judgment in selecting bands for further analysis. In any case, each putative
target must be validated using conventional ChIP assay with locus-specific
primers (step 3.4.12).
15. There are at least three reasons for the secondary amplification and digest.
First, it is useful as a screen to confirm that co-migrating bands from different
lanes actually contain the same sequence. Second, it helps to eliminate
contamination from overlapping and neighboring bands present on the
polyacrylamide gel. Restriction digestion results in a unique pattern of sub-
fragments arising from the target; sub-fragments that participate in such a
pattern will likely lead to identification of true targets (Fig. 2B). Third, the
secondary digestion facilitates the sequencing step. Because the primers used
37
for sequencing are the same as those used for amplification, they would
anneal to both ends of an intact fragment, producing two simultaneous and
thus uninterpretable sequences. Digestion of the fragments is intended to
create a sequence-able template with only one end to which a primer can
anneal. Because most of the fragments isolated after PAGE are up to 500-bp
in length, it is best to start the secondary digestion with enzymes that have a
short recognition sequence, usually 4-bp cutters. Restriction enzymes that
work well in our hands for the secondary digestion are HaeIII, MspI, RsaI
and HinfI, all of which are active in the PCR buffer. If these enzymes do not
yield sequence-able sub-fragments, digestion with a panel of additional 4-bp
cutters, or even 5- or 6-bp cutters alone or in combination, is advisable.
Again, to avoid a desalting step, it is helpful to select enzymes that are active
in the PCR buffer (a.k.a. primer extension mix).
16. Bands excised from the agarose gel must have one intact end containing the
PCR primer in order to be sequenced. Because some bands may have lost
both original ends during the secondary digest, we usually excise more than
one band from the agarose gel, with preference towards those that are only
slightly smaller than the undigested fragment. It is also advisable to excise
several replicate sub-fragments because even sequence-able sub-fragments
can become corrupted during the gel extraction, or during the sequencing step
itself. Having replicates increases the odds of obtaining a sequence.
38
17. For validation by conventional ChIP, there are at least two important
considerations. First, CD does not disclose the exact sequence to which
transcription factors bind. The discrete binding site may lie anywhere within
the AvaII fragment, or even just outside the fragment. Thus, when designing
primers for validation, it may be useful to target more than one sequence
within, or immediately surrounding, the AvaII fragment. Cognate binding
motifs found in the region can provide guidance for primer design. Second,
validation of a target disclosed by CD is most likely to occur when using the
same ChIP material that was used as input for the CD experiment. This is
due to the fact that transcription factor occupancy at some loci may be
sensitive to minor alterations in culture conditions. For this reason, it may be
wise not to use all of the original ChIP material for CD, so that the leftover
can be employed for validation. In addition, we recommend the validation of
each target in independent ChIPs, which would demonstrate that occupancy
by the transcription factor of interest occurs regardless of minor alterations to
the culture conditions.
39
Chapter 4: Identification of Novel Runx2 Targets in Osteoblasts:
Cell Type-Specific BMP-Dependent Regulation of Tram2

4.1 Introduction
The few known Runx2 target genes have been typically identified based on a
candidate approach. Promoters of genes with established roles in osteoblast biology
were screened for Runx2 cognate sites, and the functionality of these sites was
confirmed by protein-DNA and transcription assays [Javed et al., 2001; Jimenez et
al., 1999; Kern et al., 2001; Sato et al., 1998; Stock et al., 2003; Thirunavukkarasu et
al., 2000]. Obviously, such approaches cannot yield novel Runx2 target genes. Two
groups have used microarrays to identify novel Runx2-regulated genes based their
differential expression in cells from Runx2
-/-
versus wild type mice [Hecht et al.,
2007; Vaes et al., 2006]. While this approach dramatically increases the number of
candidates, it is limited to those genes spotted on the array. Moreover, expression
microarray studies are fraught with reproducibility and sensitivity problems [Irizarry
et al., 2005].

Whether a candidate Runx2 target gene emerges from archival bone literature or
from contemporary comprehensive gene expression profiling, and even if the
candidate is a direct Runx2 target, a question that cannot be effectively addressed
based on expression studies is the location of cis-acting elements responsible for the
regulation by Runx2. Traditionally, investigators searching for Runx2-binding
40
elements have focused on 5’-flanking sequences [Javed et al., 2001; Jimenez et al.,
1999; Kern et al., 2001; Sato et al., 1998; Stock et al., 2003; Thirunavukkarasu et al.,
2000]; however, there is a growing body of evidence suggesting that transcription
factors bind to other regions as well [Cawley et al., 2004; Loh et al., 2006] The task
of mapping regulatory sequences that mediate Runx2 action on target genes is
further complicated by the fact that the Runx2 consensus motif is only 6-bp long and
quite promiscuous, resulting in a large number of putative binding sites, which
greatly exceed the number of functional sites [Roca et al., 2005]. Furthermore,
sequences to which Runx2 binds in living cells may be different from the consensus,
which was defined in vitro [Meyers et al., 1993], and, Runx2 may even be recruited
to DNA indirectly via interaction with other transcription factors.

The search for transcription factor targets has been propelled in recent years by the
development of a group of novel techniques, collectively known as location analysis.
Location analysis is based on the physical interaction of transcription factors with
their target genomic loci, and thus does not suffer from the aforementioned
drawbacks. Furthermore, location analysis can discover target genes for
transcription factors even when regulation of these genes by the transcription factor
of interest occurs under conditions that are difficult to model experimentally.

Location analysis is based on chromatin immunoprecipitation (ChIP), which has
been used traditionally to test whether a candidate DNA fragment is occupied by a
41
transcription factor of interest in living cells. Because ChIP concentrates all DNA
fragments occupied by the transcription factor of interest, it can be modified to
discover novel binding regions. The simplest approach to this is cloning and
sequencing of the ChIP products. Unfortunately, ChIP-cloning is rather tedious and
yields a high false positive rate, due to the high amount of non-specifically
immunoprecipitated DNA [Weinmann et al., 2001]. Approaches that are more high-
throughput and have lower false positive rates include Serial Analysis of Binding
Elements (SABE) [Chen and Sadowski, 2005], Sequence Tag Analysis of Genomic
Enrichment (STAGE) [Kim et al., 2005], Paired-End Ditag Sequencing (ChIP-PET)
[Wei et al., 2006], Genome-Wide Mapping Technique (GMAT) [Roh et al., 2004],
Serial Analysis of Chromatin Occupancy (SACO) [Impey et al., 2004], and ChIP-
chip [Cawley et al., 2004; Horak and Snyder, 2002; Ren et al., 2000]. The major
drawbacks to these approaches are that they require large amounts of starting
material, complicated statistical analysis, and financial resources unavailable to the
average molecular biology laboratory. One approach that is less high-throughput,
but does not suffer from the aforementioned limitations is ChIP Display (CD). We
demonstrated in a pilot study that CD can discover novel targets of Runx2 in
osteoblasts [Barski and Frenkel, 2004]. In this study, we used CD to discover
additional targets, and focused on one, Tram2, which likely participates in the
execution of Runx2’s role in osteoblasts.

42
M AG M
+1,71°
M
Uncut
HaeIII
HinfI
MspI
M
Uncut
HaeIII
HinfI
MspI
M
Uncut
HaeIII
HinfI
MspI
M
Uncut
HaeIII
HinfI
MspI
+1,71°,AG +2,70°,AG +3,70°,AG +3,71°,AG C.
B. A.
AG
TG
TG M 70° 71°
+1
70° 71°
+2
70° 71°
+3
M 70° 71°
-1
70° 71°
-2
70° 71°
-3
517
396
344
298
bp
506
396
344
298
bp
396
344
298
bp
506
AG M
+2,70°
AG
TG
TG AG M
+3,70°
AG
TG
TG AG
+3,71°
AG
TG
TG
220
201
154
134
220
201
220
201
154
134
154
134
4.2 Results
ChIP Display discloses novel Runx2 genomic targets in living osteoblasts.
To discover novel Runx2 targets in osteoblasts, we employed ChIP Display (CD) of
MC3T3-E1 osteoblastic cells [Barski and Frenkel, 2004]. While our initial Runx2
CD screen consisted of the 8 reactions that each utilize a single PCR primer, we
employed in the present study all primer combinations. Furthermore, we have now
performed each PCR reaction at two annealing temperatures (see Material and
Methods). An example of the results obtained with one primer pair, (the AG and the
TG primers) is shown in Fig. 4.
Fig. 4. Representative ChIP Display results. (A) DNA was obtained from three Runx2 ChIPs (indicated +1,
+2, +3) and three IgG control precipitates (-1, -2, -3), and amplified using various combinations of eight nested
primers [Barski and Frenkel, 2004]. Products were displayed on 8% polyacrylamide gels. The one presented
here shows the results obtained with the AG/TG primer pair (see Materials and Methods). Each PCR was
performed with 70°C or 71°C as the annealing temperature, as indicated. Bands that appeared reproducible and
specific for the Runx2 ChIPs (arrowheads) were excised for identification. (B) The excised bands from panel A
were reamplified with the AG/TG primer pair used to generate the original product. In this experiment, each of
the AG and TG primers was also employed individually, as indicated. Electrophoresis in 4% agarose gels shows
that reamplification was achieved with the AG/TG primer pair, and with the AG primer alone, but not with the TG
primer. Bands indicated by arrowheads were excised and eluted from the gel. (C) The excised bands from panel
B were subjected to secondary digestion with the indicated restriction enzymes and were then resolved by 4%
agarose gel electrophoresis. The subfragments indicated by arrowheads were purified from the gel. Direct
sequencing of these subfragments was performed using the AG PCR primer. The sequence mapped to the mouse
Tram2 gene. Sizes of DNA markers (M) are indicated in base pairs.
43

Table III depicts the position of seven Runx2-occupied regions discovered in this
study and the nearby genes. Runx2 occupancy at each of these regions was
confirmed by conventional ChIP assay using locus-specific primers in at least one
independent experiment (see Fig. 5 for example). Six of the newly discovered
Runx2 targets are found within 1 kb of known genes. These include: (i) Tram2,
which has been implicated in collagen biosynthesis [Stefanovic et al., 2004]; (ii)
Lnx2, a potential modulator of Notch signaling [Rice et al., 2001]; (iii) Tnfrsf12a,
encoding a member of the TNF receptor superfamily [Wiley and Winkles, 2003];
(iv) Abcc8, which encodes the regulating

subunit of the K
+
ATP
channel [Bryan et al.,
2007]; (v) Amdhd2, which contains a predicted amidohydrolase domain
(www.ensembl.org); and (vi) Gm1082, a gene whose protein product has not yet
been characterized. One of the seven Runx2-occupied regions was found over 50 kb
from the nearest annotated transcript, which happens to be a non-coding RNA (Table
III).
Table III. Runx2-occupied genomic sites disclosed in the present study
Primer
Pair
1

Chromosome No. &
(Absolute Position)
2

Nearest Annotated Transcript
Positional Relationship To
Nearest Gene
TG,AG 1 (21064103-346)
Translocating chain-associating
membrane protein 2 (Tram2)
Straddles boundary between
first exon and intron
TC,TT 5 (147386706-7208) Ligand of numb-protein X 2 (Lnx2)
5’ flanking, first exon, part of
first intron
AC,AG 7 (30230441-697) Gene model 1082 (Gm1082) Completely within exon 12

44
Table III. (Continued)

Primer
Pair
1

Chromosome No. &
(Absolute Position)
2

Nearest Annotated Transcript
Positional Relationship To
Nearest Gene
TA,AT 7 (45975389-956)
ATP-binding cassette, sub-family C
(CFTR/MRP), member 8 (Abcc8)
Completely within intron 33
TT,AT 17 (4581782-2093) U6 snRNA
52.409 kb upstream from
transcription start site
TT,TT 17 (23405050-489)
Tumor necrosis factor receptor
superfamily, member 12a
(Tnfrsf12a)
5’ flanking & part of first exon
AT,AA 17 (23882051-423)
Amidohydrolase domain containing
2 (Amdhd2)
1.032 kb downstream from 3’
end of exon 11

1
Primer pair indicates which 2 of the 8 possible oligos were used to amplify the digested, ligated ChIP DNA (see
Materials and Methods).
2
Absolute position refers to the chromosomal location of AvaII-digestible fragments that were picked up by the
CD screen. Annotation is based on the NCBI m36 mouse assembly.

Interestingly, only 3 of those fragments found near known genes are positioned in
the vicinity of the respective transcription start site (Table III). One of these
encompasses the exon1/intron1 boundary of Tram2 (Fig. 5A), a gene necessary for
proper synthesis and secretion of collagen type I [Stefanovic et al., 2004]. As
depicted in Fig. 5A, a perfect Runx2 consensus motif is present 210-bp downstream
of the AvaII restriction fragment, which was excised from the CD acrylamide gel
(Fig. 7). The conventional ChIP assay to confirm Runx2 occupancy at the Tram2
locus was performed using primers flanking the Runx2 motif (Fig. 5B). The
rationale for this design is that fragments generated during ChIP are in the order of
500-1000 bps; thus, precipitation of the 243-bp AvaII fragment likely occurred due to
binding of Runx2 to its adjacent cognate motif, not to any sequence within the 243-
45
+483 +231 +1
ACACCA
Restriction
Fragment
OC Ins Tram2
ChIP
Genomic
Tram2
+30 +273
A. B.
bp AvaII fragment (Barski et al, 2007). Be that as it may, the robust signal obtained
in the conventional ChIP assay (Fig. 5B) leaves no doubt that Runx2 occupies this
locus, although, like any ChIP assay, it does not allow precise mapping. Given the
role of Tram2 in collagen synthesis, and since collagen type I is the most abundant
organic component of the bone extracellular matrix, we hypothesized that Runx2
might be positively regulating Tram2 expression.

Fig. 5. Runx2 interacts with Tram2. (A) A restriction fragment from the CD screen (bounded by vertical
arrows) is shown in its genomic context, where it overlaps the first exon of Tram2 (black rectangle). Positions
are numbered relative to the Tram2 transcription start site (+1). Primers used for conventional ChIP assay
(horizontal arrows) flank a perfect Runx2 consensus motif (ACACCA). (B) Conventional Runx2 ChIP assay was
performed with locus-specific primers flanking the Tram2 Runx2 motif shown in panel A. The region containing
the Runx2 element from the murine osteocalcin (OC) promoter [Banerjee et al., 1997; Ducy and Karsenty, 1995]
was amplified as a positive control, and a region from the insulin promoter (Ins) served as negative control. The
sequences of the primers used to amplify the Tram2, OC and Ins regions are provided in Table I. Increasing
amounts of genomic DNA were also amplified with the same primers under the same conditions, to show that the
PCR was performed within a dynamic range. All PCR products were visualized on a 1% agarose gel stained with
ethidium bromide.

Runx2 inhibits Tram2 expression in non-osteoblasts
The role of Runx2 in the regulation of Tram2 gene expression was initially addressed
using mouse primary skeletal myoblasts retrovirally infected with Runx2. As shown
in Fig. 6A, these cells expressed osteocalcin, a classical Runx2 target gene at levels
100-fold higher than unmodified myoblasts. Surprisingly, Tram2 mRNA was 4.5-
46
fold lower in the Runx2-infected cells. These results suggest that Runx2 inhibits,
rather than stimulates expression of Tram2. The ability of Runx2 to stimulate and
repress different genes, even in the same cells, has been previously described
[Gersbach et al., 2006; Javed et al., 2001] and is attributable to the presence of both
activation and repression domains in Runx2 [Westendorf, 2006]. We further
investigated the influence of Runx2 on Tram2 in C3H10T1/2 mouse embryonic
fibroblasts transiently transfected with Runx2. As shown in Fig. 6B, Tram2 mRNA
levels in Runx2-transfected cells were approximately half of those observed in
control cells. Thus, Tram2 expression was inhibited in two different cell lines both
after prolonged expression of Runx2 (Fig. 6A) and 48 hours following transient
transfection of Runx2 (Fig. 6B).
47
0
0.5
1
1.5
2
0
1
2
3
4
5
0
0.5
1
1.5
2
2.5
Runx2 OC Tram2
Primary Skeletal
Myoblasts
CH310T1/2
CH310T1/2
+rhBMP2
A.
B.
C.
Relative mRNA Levels (corrected for rpL10A)
*
*
*
*
* *
*
*
Fig. 6. Runx2 inhibits Tram2 expression in non-osteoblasts. (A) Primary skeletal myoblasts retrovirally
transduced with Runx2 (black bars) and unmodified myoblasts (white bars) were cultured for 7 days, after which
total RNA was extracted and converted to cDNA. Runx2, Osteocalcin (OC), and Tram2 expression levels were
then determined by qPCR. Bars represent the Mean ± SEM (n=3) of values corrected for the expression of
ribosomal protein L10A, whose mRNA levels were not affected by Runx2 over-expression. (B) C3H10T1/2
mouse embryonic fibroblasts were cultured for 48 hours after transient transfection with Runx2 expression
construct (black bars) or empty vector (white bars). Total RNA was collected, reverse-transcribed, and analyzed
as described for panel A. Transfection efficiency was approximately 80%, as indicated by FACS analysis of
parallel cultures transfected with pGFP-C1 (Clontech). (C) Same as in panel B, except 300 ng/ml rhBMP-2 was
administered for 24 hours prior to harvest. *, p ≤ 0.05

Tram2 is positively regulated along with osteocalcin late during BMP-2-induced
osteoblastic differentiation of C3H10T1/2 mouse embryonic fibroblasts
BMP signaling plays a major role in osteoblast differentiation, and also modifies
Runx2 activity [Chen et al., 2004; Guicheux et al., 2003; Jeon et al., 2006; Lee et al.,
2000; Xiao et al., 2002]. We therefore measured Tram2 mRNA levels in
48
C3H10T1/2 cells that were both transfected with Runx2 and treated with
recombinant human bone morphogenetic protein 2 (rhBMP-2). As shown in Fig. 6C,
the inhibitory effect of Runx2 on Tram2 gene expression was almost completely
alleviated in the presence of rhBMP-2. Thus, the Runx2-mediated inhibition of
Tram2 expression was strongest in primary skeletal myoblasts, weaker in naive
untreated pluripotent C3H10T1/2 cells, and nearly absent in rhBMP-2-treated
C3H10T1/2 cells.

Given the ability of rhBMP-2 to modify the effect of exogenous Runx2 on Tram2
expression in transiently transfected C3H10T1/2 cells, we next investigated whether
there was any evidence that endogenous Runx2 contributes to the regulation of
Tram2 expression in long-term C3H10T1/2 cultures, treated or untreated with
rhBMP-2. As shown in Fig. 7A, Runx2 mRNA was detectable in confluent cultures
(day 0), and was thereafter up-regulated in both untreated and BMP2-treated cells.
After two days of treatment, Runx2 mRNA continued to increase in the BMP2-
treated but not in the untreated cells. Interestingly, expression of OC, a classical
Runx2 target gene, was hardly detectable in either untreated or BMP2-treated
cultures on day 4, when Runx2 expression was maximal. Only later did OC mRNA
increase and this was only observed in the BMP-2-treated cultures (Fig. 7B). Tram2
expression in untreated cultures increased between day 0 and 1, in parallel to the
increase in Runx2 expression. Later, when BMP-2 stimulated Runx2 expression,
Tram2 responded similar to OC. Specifically, it was not influenced by BMP-2 until
49
0.0
0.5
1.0
1.5
2.0
Runx2 mRNA
0
5
10
15
20
25
30
35
Osteocalcin mRNA
0.0
0.5
1.0
1.5
2.0
2.5
Tram2 mRNA
2 0468
Days After Treatment
CH310T1/2
A.
B.
C.
*
* *
*
*
day 4, even though by this point Runx2 mRNA levels had already peaked.
Thereafter, however, Tram2 expression was higher in the presence of BMP-2 (day
8), like that of OC mRNA. The similarity between Tram2 and OC expression
observed after Runx2 is induced by BMP-2 treatment suggests that Runx2 positively
regulates Tram2 transcription during the latter stages of osteogenic differentiation in
the C3H10T1/2 culture model.

Fig. 7. Upregulation of Tram2 mRNA in BMP-2 treated embryonic fibroblasts. CH310T1/2 cells were
cultured in the presence (closed circles) or absence (open circles) of 300 ng/ml rhBMP-2, and total RNA was
collected at the indicated times. Relative mRNA levels were determined by RT-qPCR for (A) Runx2, (B)
Osteocalcin (OC), and (C) Tram2. All data points are the Means ± SEM (n=3) of values corrected for the
expression of ribosomal protein L10A, whose mRNA was not affected by BMP-2 treatment. *, p ≤ 0.05
50
Tram2 is up-regulated soon after BMP-2 treatment of ST2 bone marrow-derived
stromal cells
We further investigated Tram2 expression during rhBMP2-mediated osteogenic
differentiation of ST2 bone marrow-derived stromal cell cultures. As shown in Fig.
8A, Runx2 mRNA was detectable in confluent ST2 cultures (day 0), and was
thereafter increased in both untreated and BMP2-treated cells. However, after two
days of treatment, Runx2 mRNA was expressed at higher levels in the cells treated
with BMP-2. At this time point, OC mRNA was still hardly detectable. OC
expression commenced after day 2, and was remarkably stronger in the BMP-2-
treated cultures (Fig. 8B). Interestingly, Tram2 mRNA, which, unlike OC, was
already detected on day 0, responded to BMP2 treatment as early as day 2, at which
Runx2 was first upregulated in the treated compared to untreated cultures (Fig. 8C).
These results are consistent with Runx2 controlling Tram2 gene expression; the
earlier response of Tram 2 compared to OC likely reflects the inactivity of positive
regulatory mechanisms, or the activity of negative regulatory mechanisms, which
result in lack of basal OC expression on day 2.
51
0.0
0.5
1.0
1.5
2.0
2.5
Runx2 mRNA
0
2
4
6
8
10
Osteocalcin mRNA
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Tram2 mRNA
2 0468
Days After Treatment
ST2
A.
B.
C.
*
*
0.0
0.5
1.0
1.5
2.0
2.5
Runx2 mRNA
0
2
4
6
8
10
Osteocalcin mRNA
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Tram2 mRNA
2 0468
Days After Treatment
ST2
A.
B.
C.
*
*

Fig. 8. Early upregulation of Tram2 mRNA in rhBMP-2 treated bone marrow-derived stromal cells. ST2
cells were cultured and treated following the same protocol as in Fig. 7. Gene expression was analyzed and the
data was plotted as in Fig. 7 for (A) Runx2, (B) Osteocalcin (OC), and (C) Tram2. Closed circles, rhBMP-2 (300
ng/ml); open circles, control. Mean ± SEM (n=3). *, p ≤ 0.05.
52
Runx2, OC and Tram2 are transiently upregulated in BMP-2-treated C2C12
myoblasts
C2C12 myoblasts undergo transdifferentiation and acquire an osteoblast-like
phenotype in response to BMP-2 treatment [Katagiri et al., 1994]. This
transdifferentiation entails the induction and activation of Runx2 and Runx2-target
genes [Lee et al., 2000]. We observed transient upregulation of Runx2 in post-
confluent C2C12 cultures even in the absence of rhBMP-2 (Fig. 9A); however, this
upregulation was brief, and was not accompanied by an induction of OC mRNA
(Fig. 9B). In contrast, the upregulation of Runx2 mRNA in BMP-2-treated cells was
sustained for several days, and was followed by similar upregulation of OC mRNA.
The Tram2 expression profile in the BMP-2-treated cultures paralleled that of OC,
with both mRNAs reaching maximum levels on day 2 (Fig. 9B, C). However,
reminiscent of the ST2 cells, the C2C12 cultures already expressed Tram2, but not
OC, on day 0 (Fig. 9B, C). By day 2, Tram2 mRNA in untreated cultures was
slightly upregulated in parallel to the brief expression of Runx2. As compared to the
untreated cultures, the BMP2-treated cultures had 1.5-fold more Tram2 mRNA on
day 2, and 2.2-fold more Tram2 mRNA on day 4, as basal levels declined (Fig. 9C).
53
0
1
2
3
4
5
6
7
Runx2 mRNA
0
10
20
30
40
50
60
Osteocalcin mRNA
0
2
4
6
8
10
12
14
Tram2 mRNA
2 0468
Days After Treatment
C2C12
A.
B.
C.
*
*
*
*
*

Fig. 9. Transient upregulation of Tram2 mRNA in rhBMP-2 treated C2C12 pre-myoblasts. Gene
expression in C2C12 cell cultures was measured under the same conditions described in Figs. 7 and 8. Each data
point is the Mean expression ± SEM (n=3) of (A) Runx2, (B) Osteocalcin (OC), and (C) Tram2. Closed circles,
rhBMP-2 (300 ng/ml); open circles, control. *, p ≤ 0.05.

54
Tram2 expression in MC3T3-E1 osteoblastic cells is regulated by BMP-2 in a
developmental stage-specific manner
We finally investigated the expression pattern of Tram2 in MC3T3-E1
preosteoblasts, in which the physical interaction of this gene with Runx2 was
originally discovered (Figs. 1 & 2). In BMP-2 treated cultures, Tram2 mRNA levels
behave like those of OC after Runx2 activation. That is, they steadily increase from
day 1 throughout the remainder of the time course (Figs. 7B & C). Interestingly,
Tram2 expression does not parallel that of Runx2 in the absence of BMP-2.
Between day 0 and 1, there is a dramatic increase in Tram2 mRNA with an
antiparallel decrease in Runx2 mRNA; thereafter, Tram2 mRNA levels decrease as
Runx2 levels increase (Fig. 10A, C). The inverse relationship between Runx2 and
Tram2 in untreated MC3T3-E1 cells is reminiscent of the results with primary
myoblasts (Fig. 6A) and untreated C3H10T1/2 cells (Fig. 6B). In the same untreated
MC3T3-E1 cells, OC mRNA is undetectable until day 1 (Fig. 10B). Thereafter, it
increases in parallel to Runx2 mRNA (Fig. 10A, B), suggesting differential
regulation of OC versus Tram2 transcription by Runx2 in untreated MC3T3-E1 cells.

55
0.0
0.5
1.0
1.5
2.0
2.5
Runx2 mRNA
0
1
2
3
4
5
6
7
Osteocalcin mRNA
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
Tram2 mRNA
2 0468
Days After Treatment
MC3T3-E1
A.
B.
C.
*
*
*
*
*
*
0.0
0.5
1.0
1.5
2.0
2.5
Runx2 mRNA
0
1
2
3
4
5
6
7
Osteocalcin mRNA
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
Tram2 mRNA
2 0468
Days After Treatment
MC3T3-E1
A.
B.
C.
*
*
*
*
*
*

Fig. 10. Tram2 expression in osteoblastic cells is regulated by BMP-2 in a developmental stage-specific
manner. MC3T3-E1 cells were cultured in the presence (closed circles) or absence (open circles) of 300 ng/ml
rhBMP-2, and total RNA was collected at the indicated times. Relative mRNA levels were determined by RT-
qPCR for (A) Runx2, (B) Osteocalcin (OC), and (C) Tram2. Each data point represents the Mean ± SEM (n=3)
of values corrected for the expression of ribosomal protein L10A, whose mRNA was not affected by BMP-2
treatment. *, p ≤ 0.05

56
4.3 Discussion
Using CD of MC3T3-E1 osteoblasts, we discovered seven Runx2 genomic sites that
were not previously known to be occupied by Runx2 in living cells. Unlike location
analyses of other transcription factors, where many occupied sites were found tens of
kb away from nearby known genes, our results suggest that Runx2 may occupy
regions that tend to be associated with transcribed sequences. All but one of the 12
Runx2 occupied regions in our present and previous study were either within
annotated genes or up to 1-kb from either end (Table III and [Barski and Frenkel,
2004]). Of the 11 intragenic Runx2 sites, 6 were either within or overlapping the
first intron. Albeit limited in scope, these results also highlight the evolving concept
that cis-acting regulatory elements are not necessarily confined to 5’ flanking gene
sequences [Bieda et al., 2006; Cawley et al., 2004], but suggest that Runx2-occupied
regions are frequently found within or 3’ to transcribed sequences. One of our
Runx2 hits was in the first intron of Tram2, immediately downstream of the first,
231-bp long exon. The evidence that Tram2 is regulated by Runx2 include: 1)
Tram2 mRNA is downregulated in response to ectopic Runx2 expression in primary
skeletal myoblasts and C3H10T1/2 embryonic fibroblasts; and 2) Tram2 expression
parallels that of osteocalcin in the C3H10T1/2, ST2, C2C12, and MC3T3-E1 cell
lines. The apparently opposing effects of Runx2 on Tram2 expression in the
different experimental systems may represent alterations in Runx2 activity at the
Tram2 locus as a function of cell type or developmental stage (see below). Be that
as it may, given the role of Tram2 in type I collagen synthesis [Stefanovic et al.,
57
2004], and given the role of type I collagen in bone, it is likely that Tram2 belongs to
the machinery engaged by Runx2 in promoting osteoblast differentiation and
function.

Cell Type-dependent and Gene-Specific Transcriptional Regulation by Runx2
Runx2 has been shown to function as a transcriptional activator or repressor
depending on the cellular milieu and promoter context [Gersbach et al., 2006; Javed
et al., 2001]. We provide further evidence of this by showing that Runx2 is
associated with either Tram2 repression or activation depending on the cell type and
culture conditions. Further, in some instances, e.g., primary myoblasts, Runx2
repressed Tram2 while activating osteocalcin expression. These phenomena are
likely explained by the involvement of co-regulators that either stimulate or repress
Runx2 transcriptional activity. Transcription factors that positively modify Runx2’s
activity include Cbf β [Yoshida et al., 2002], C/EBP β and C/EBP δ [Gutierrez et al.,
2002], ETS1 [Sato et al., 1998], Menin [Sowa et al., 2004], Smad1 [Zhang et al.,
2000], Smad5 [Nishimura et al., 2002], Grg5 [Wang et al., 2004], Rb [Thomas et al.,
2001], TAZ [Cui et al., 2003], and p204 [Liu et al., 2005], whereas negative co-
regulators of Runx2’s activity include C/EBP δ [McCarthy et al., 2000], Dlx3
[Hassan et al., 2004], Msx2 [Shirakabe et al., 2001], PPAR γ [Jeon et al., 2003],
HDAC4 [Vega et al., 2004], HDAC3 [Schroeder et al., 2004], Stat1 [Kim et al.,
2003], Twist [Bialek et al., 2004], Yes [Zaidi et al., 2004], TLE [Javed et al., 2000],
and Smad3 [Kang et al., 2005]. Thus, the ability of Runx2 to exert different effects
58
on the same gene in different cells may be due to the differential expression of co-
regulators among cell types. Likewise, the ability of Runx2 to exert opposing effects
on different genes in the same cell may be due to preferential recruitment of
coactivators or corepressors, depending on the sequences in the vicinity of the Runx2
binding site. Finally, it is conceivable that particular genes may be initially
stimulated and then repressed by Runx2 during development of the osteoblast
phenotype. The biphasic expression pattern of Tram2 as a function of time in
MC3T3-E1 and C2C12 cultures may reflect such transition in Runx2 activity at this
locus.

BMP-Dependent Transcriptional Regulation by Runx2
BMPs are potent osteogenic agents [Chen et al., 2004], in part due to their ability to
up-regulate Runx2 expression [Lee et al., 1999]. While BMP-2 treatment does not
increase the interaction of Runx2 with the OC promoter in MC3T3-E1 cells (our
unpublished ChIP results), several mechanisms have been reported, through which
BMPs post-translationally enhance Runx2 transcriptional activity. For example,
p300-mediated acetylation of Runx2 is induced by BMP2, and this prevents Smurf-
mediated degradation of Runx2 [Jeon et al., 2006]. Also, Runx2 is a substrate for
MAPK [Xiao et al., 2002], which in turn can be activated by BMP signaling
[Guicheux et al., 2003]. Finally, Smad-1, -5 and -8, which are activated by BMPs,
can associate with and augment Runx2-mediated transcription [Lee et al., 2000].
Consistent with these notions, we show here that (i) the Runx2-mediated repression
59
of Tram2 in C3H10T1/2 cells was alleviated by the addition of rhBMP-2, (ii)
rhBMP-2 enhanced Runx2 and Tram2 levels compared to untreated cells in the four
cell lines tested; and (iii) the relationship between Tram2 and Runx2 expression in
MC3T3-E1 cells was completely reversed upon treatment with rhBMP-2. A likely
interpretation of these results is that rhBMP-2 altered the Runx2-mediated regulation
of Tram2. However, we cannot rule out BMP-dependent Runx2-indepedent
regulation of Tram2 expression.

Potential Novel Mediators of Osteoblast Development and Function
The present study identifies translocon-associated membrane protein 2 (Tram2) as a
Runx2 target gene. Tram2 was originally identified in a microarray screen for genes
stimulated in a stellate cell culture model of cirrhosis [Stefanovic et al., 2004]. Most
of the Tram 2 domains are highly homologous to TRAM, an integral component of
the protein translocon found on the endoplasmic reticulum (ER). The C-terminus of
Tram2, which shares only 15% homology with Tram, was found to interact with
Serca2b, the Ca
2+
pump

of the ER [Stefanovic et al., 2004], and over-expression of a
Tram2 deletion mutant lacking the C-terminus resulted in decreased collagen
accumulation. Since depletion of ER Ca
2+
stores had a similar effect, the authors
proposed that

Tram2 may recruit Serca2b to the

translocon, leading to increased local
Ca
2+
concentration

and stimulation of chaperone-mediated folding of triple helical

procollagen [Stefanovic et al., 2004]. Like activated stellate cells, osteoblasts
produce large amounts of type I collagen. Runx2 may promote osteoblast
60
differentiation and function by enhancing the transcription of not only type I
collagen [Kern et al., 2001], but also genes like Tram2, which are necessary for
effective post-translational processing of this macromolecule.

Other Runx2 target genes identified in this study may also help explain the
importance of Runx2 in skeletogenesis. Numb is an inhibitor of Notch signaling in
Drosophila [Guo et al., 1996; Spana and Doe, 1996], and likely in mammalian cells
as well [Petersen et al., 2006], and Notch signaling plays an important role in
osteoblast differentiation [Deregowski et al., 2006]. We have found that Runx2
occupies the gene encoding Ligand of Numb-related protein 2 (Lnx2), a scaffolding
molecule for mammalian Numb and Numblike [Rice et al., 2001]. It is conceivable
that regulation of Lnx2 by Runx2 contributes to the osteogenic program.

Another gene that was associated with Runx2 in living MC3T3-E1 osteoblastic cells
is Tumor necrosis factor receptor super-family 12a (TNFRsf12a). Stimulation of this
TNF-like receptor in MC3T3-E1 cells results in concomitant inhibition of osteoblast
differentiation and induction of the osteoclast activators RANKL and RANTES
[Ando et al., 2006]. Regulation of TNFRsf12a by Runx2 might therefore play a role
in modulating the activities of both osteoblasts and osteoclasts.

In summary, we used CD to identify seven genomic loci that are occupied by Runx2
in developing MC3T3-E1 pre-osteoblasts. Expression of one of these genes, Tram2,
61
is regulated by Runx2 in a manner dependent upon cell type and the status of the
BMP signaling pathway. Further investigation is warranted to delineate the potential
roles of Tram2 and the other novel Runx2 targets in bone biology, as well as to
expand the repertoire of Runx2-regulated genes in osteoblasts.
62
Chapter 5: An Expression-Based Approach for Identifying Novel
Runx2 Targets

Measuring gene expression changes in response to Runx2 over-expression is a
complementary approach to location analysis that can potentially lead to the
identification of novel Runx2 targets. Ideally, it should be performed in the same
cells used for location analysis; however, we were unable to accomplish this for
technical reasons. For example, although MC3T3-E1 cells were suitable for location
analysis, they were unsuitable for the over-expression of Runx2. This was evident
from experiments where exogenous Runx2 was introduced into MC3T3-E1 cells by
transient transfection of a Runx2 expression vector, but without any accompanying
changes in gene expression, even for well-known targets such as osteocalcin (Fig.
11). The underlying cause of this was most likely poor transfection efficiency,
which was estimated to be about 50% by FACs analysis of MC3T3-E1 cells that
were transfected with a GFP-expression vector (data not shown).

63

Fig. 11. Effect of exogenous Runx2 on gene expression in MC3T3-E1 cells.

Osteopontin
0
5
10
15
20
25
30
Bone Sialoprotein
0
10
20
30
40
Mmp9
0
5
10
15
20
25
30
Mmp13
0
5
10
15
20
OC (2nd Analysis)
0
10
20
30
40
Osterix
0
5
10
15
20
25
1231231231 23123123
Runx2 CMV Runx2 CMV Runx2 CMV
24 Hours 48 Hours 72 Hours
Runx2
0
10
20
30
40
50
Osteocalcin
0
5
10
15
20
25
30
Type I Collagen
0
5
10
15
20
25
30
Rpl10a
0
2
4
6
8
10
12
14
16
Gapdh
0
5
10
15
20
B-actin
0
5
10
15
20
12 312 312 312 312 312 3
Runx2 CMV Runx2 CMV Runx2 CMV
24 Hours 48 Hours 72 Hours
A.
B.
C.
D.
E.
F.
G.
H.
I.
J.
K.
L.
64
Another cell line that we have used extensively, C3H10T1/2 embryonic fibroblasts,
are much more suitable for over-expression of Runx2; however, they are not suitable
for location analysis. Evidence for the former comes from our previous experiments
with Tram2. Evidence for the latter came out of our attempts to perform ChIP in
C3H10T1/2 cells transiently-transfected with Runx2 expression vector. We first
collected material from these cells by cross-linking six 100-mm plates for 10 minutes
each and sonicating them each on ice for 1 minute (15-second pulses). After
centrifugation, we removed an aliquot and snap-froze the remainder. The aliquot
was incubated at 65°C overnight to reverse the cross-links. The samples were then
digested with Proteinase K, extracted with phenol/chloroform, ethanol precipitated,
re-suspended and digested with RNase. Thirty percent (10-ul) of the final product
was electrophoresed on a 1% agarose gel stained with EtBr. All three of the samples
had fragments that ranged in size from anywhere between 5 and 0.3 Kb (Fig. 12A).
Although it was not clear exactly where the average fragment size lies, it did not
appear to be 500bp or lower. Thus, we thawed the remainder of the samples and
sonicated samples 2 and 3 for an additional 1 and 2 minutes on ice, respectively. We
then removed a 10-ul aliquot from all three samples and reversed the cross-links
overnight at 65°C. The remainder was snap-frozen and stored at -80°C. The three
samples were then subjected to Proteinase K digestion, phenol/chloroform
extraction, ethanol precipitation and RNase digestion, followed by gel
electrophoresis on 1% agarose stained with EtBr. It appeared that the average
fragment size had been reduced as a result of further sonication (Fig. 12B). This was
65
true of the sample sonicated for 2 minutes and more so for the sample sonicated for 3
minutes. Indeed, it appeared that the average fragment size for these samples was
below the 500 bp mark, even though there were still some large molecular weight
fragments present.

Fig. 12. Gel analysis of chromatin isolated from C3H10T1/2 cells.

Therefore, we thawed out the samples again and removed an aliquot of 200-ul from
each for IP (100-ul for Runx2 IP, 100-ul for IgG), and an aliquot of 10-ul for input
DNA. The samples were processed to completion, and the enrichment levels were
measured by real-time PCR. Oddly, there was appreciable enrichment in the IgG
samples only (Fig. 13). This was true at least for the samples that were sonicated for
2 and 3 minutes, with longer sonication corresponding to greater enrichment. For the
sample that was sonicated for only 1 minute, there was no significant enrichment in
1.6
2.0
3.0
MW
(Kb)
4.0
5.0
0.5
0.4
0.3
0.2
2’ 3’ 1’
1.6
2.0
3.0
MW
(Kb)
4.0
5.0
0.5
0.4
0.3
0.2
#1 #2 #3
1 minute
A. B.
66
either the Runx2 or IgG sample. There are at least two possible explanations for
these observations. First, the enrichment seen in the IgG samples could be spurious.
Second, the Runx2 and IgG samples may have been switched by mistake.

Fig. 13. ChIP analysis of Runx2 occupancy in C3H10T1/2 cells.

To discern between these possibilities, we plated twelve 100-mm dishes of
C3H10T1/2 cells for transfection with Runx2 expression vector, transfected each
plate with 24-ug of vector, and cross-linked the cells after 48 hours. Six of the plates
were cross-linked for 5 minutes, and the remaining six were cross-linked for 10
minutes. All the reactions were quenched with 125mM glycine. Of the resulting 3
samples derived from cells that had been cross-linked for 5 minutes, we sonicated
one for 1 minute on ice (30-second pulses), one for 2 minutes, and one for 3 minutes.
We did likewise for the three samples derived from cells that had been cross-linked
for 10 minutes. We then removed a 10-ul aliquot from each of these for fragment
size analysis, and immunoprecipitated the remainder with either Runx2-specific
antibody or IgG. After incubation overnight at 65°C to reverse the cross-links, the
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
Runx2 IgG Runx2 IgG Runx2 IgG
1' 2' 3'
Percentage Input
Osteocalcin
Insulin
67
aliquots were proteinase K-digested, phenol/chloroform-extracted, ethanol-
precipitated, and RNase-treated. They were then electrophoresed on 1% agarose
stained with EtBr. Visualization with UV light revealed that most (if not all) the
samples contained fragments that were of an acceptable size range (Fig. 14A).
Additionally, unbound DNA from the immunoprecipitation was washed away, and
the crosslinks were reversed overnight at 65°C. After cross-link reversal, the
samples were digested with Proteinase K, extracted with phenol/chloroform,
precipitated with ethanol, and analyzed by real-time PCR. Most of the significant
enrichment in this experiment was seen in the IgG samples (Fig. 14B), suggesting
that the OSE2 element of the osteocalcin promoter was unoccupied by Runx2 in
these cells, despite the fact that osteocalcin mRNA was up-regulated.

Fig. 14. Gel and ChIP analysis of chromatin collected from C3H10T1/2 cells.

Thus, although it would be ideal to perform location and expression analysis in the
same cell line, technical challenges forced us to abandon this plan. Instead, we
elected to proceed with expression analysis in C3H10T1/2 cells and with location
1.6
2.0
3.0
MW
(K b )
4.0
5.0
0.5
0.4
0.3
10’ 5’
Sonication:
Cross-linking :
1’ 3’ 2’ 1’ 3’ 2’
0.6
0.4
2.0
0.3
MW
(Kb)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
1' 2' 3' 1' 2' 3'
5' 10'
Percentage Input
Osteocalcin
Insulin
A.
B.
68
1.6
2.0
3.0
MW
(Kb)
4.0
5.0
6.0
#2 #1 #3 #1 #2 #3
+Runx2 -Runx2
analysis in MC3T3-E1 cells. For expression analysis, we collected RNA from
C3H10T1/2 cells that had been transfected (or not) with Runx2 expression vector 48
hours earlier. Transfection efficiency (~80%) was assessed by FACS analysis of 3
wells transfected with GFP expression vector and 3 wells transfected with an
equimolar amount of pcDNA 3.0 (data not shown). We took 640-ug of RNA from
each sample and analyzed half of this (320-ug) by agarose gel electrophoresis (Fig.
15).

Fig. 15. Gel analysis of RNA collected from C3H10T1/2 cells.

The other half was reverse-transcribed to make 20-ul of cDNA. Expression levels
were then measured in each of the cDNA samples for Runx2, Osteocalcin, Bone
Sialoprotein, Osteopontin, Mmp9, Mmp13, Rpl10A, and Gapdh. Runx2 levels were
significantly higher in the Runx2-transfected cells, as expected (Fig. 16B). In
addition, osteocalcin levels were significantly higher in these cells as well. This is in
accordance with results obtained from the previous Runx2-over-expression
69
experiment done in C3H10T1/2 cells (Pregizer, et al. 2007). Interestingly, there was
no significant difference in the amount of expression for either of Osteopontin,
Mmp9 or Mmp13 (Fig. 16 E,F,G). Measurement of Bone sialoprotein expression
was unsuccessful in this experiment. Gapdh and Rpl10A expression was similar in
both Runx2-transfected cells and controls, as expected (Fig. 16A, D). In nearly all
cases, expression levels for the #2 control were lower than that of #1 and #3, an
observation which may be attributable to a difference in starting amount, as gel
analysis of the RNA would seem to indicate (see previous figure).

70

Fig. 16. RT-qPCR analysis of RNA collected from C3H10T1/2 cells transfected with Runx2.
Runx2
0
1
2
3
4
Osteocalcin
0
1
2
3
Rpl10a
0
1
2
Gapdh
0
1
2
Osteopontin
0
1
2
3
Mmp9
0
1
2
1 231 23
Runx2 CMV
Mmp13
0
1
2
3
A.
B.
C.
D.
E.
F.
G.
71
Given that we were able to successfully introduce exogenous Runx2 into C3H10T1/2
cells and measure its effects on known target genes, we next wanted to measure its
effects on global gene expression in this system, via expression array analysis, in
order to identify novel Runx2 target genes. To this end, we amplified and labeled
RNA collected from cells with exogenous Runx2 and those without. This material
was then hybridized to expression arrays containing probes specific for all known
transcribed sequences in mice. The arrays were scanned and the signals representing
hybridization intensity were collected for each probe. The average of 3 biological
replicates was calculated, and the averages obtained for cells transfected with Runx2
were divided by the averages obtained for cells that were transfected with a control
vector. The resulting ratio represented the fold-change in expression for a given
transcript in response to exogenous Runx2. The genes for which this fold-change is
greater than or equal to 1.5, and for which the p-value is less than 0.005 are shown in
Table IV.

Table IV. Genes with altered expression in the presence of exogenous Runx2
Entrez Gene Gene Symbol Ratio p-value
12393 Runx2 5.33 0.00007
58185 Rsad2 4.75 0.00026
15945 Cxcl10 4.61 0.00101
12393 Runx2 4.49 0.00050
15061 H28 4.45 0.00015
20296 Ccl2 4.29 0.00257
58185 Rsad2 4.19 0.00615
17857 Mx1 4.14 0.00301
66607 Ms4a4d 4.01 0.00410
58185 Rsad2 3.50 0.00572
20304 Ccl5 3.43 0.00050
15958 Ifit2 3.25 0.01821
72
Table IV. (continued)

Entrez Gene Gene Symbol Ratio p-value
15061 H28 3.19 0.00114
57444 Isg20 2.97 0.03062
20306 Ccl7 2.90 0.00023
107569 Nt5c3 2.87 0.00283
67138 Herc5 2.86 0.00475
67138 Herc5 2.64 0.02093
14468 Gbp1 2.61 0.00882
74568 Mlkl 2.57 0.04918
24110 Usp18 2.49 0.00392
22169 Tyki 2.48 0.01687
67138 Herc5 2.34 0.00564
16912 Psmb9 2.27 0.00925
60440 Iigp1 2.26 0.02625
19171 Psmb10 2.20 0.02310
21822 Tgtp 2.20 0.00559
16068 Il18bp 2.18 0.01625
72119 Tpx2 2.13 0.01411
217203 Tmem106a 2.12 0.00394
11815 Apod 2.11 0.00664
67138 Herc5 2.03 0.00105
14469 Gbp2 2.03 0.01287
20723 Serpinb9 2.02 0.00605
16913 Psmb8 2.00 0.00628
17082 Il1rl1 2.00 0.01795
142688 Asb13 2.00 0.03182
56066 Cxcl11 1.98 0.01243
234311 BC013672 1.98 0.03217
60440 Iigp1 1.97 0.01526
14469 Gbp2 1.95 0.01462
433470 AA467197 1.95 0.01266
58203 Zbp1 1.93 0.00663
216549 Aftph 1.92 0.00286
98999 AI481105 1.89 0.03962
64685 Nmi 1.88 0.02159
142980 Tlr3 1.84 0.02385
102084 AI451557 1.83 0.00137
269336 Ccdc32 1.83 0.02907
14972 H2-K1 1.82 0.00214
22040 Trex1 1.82 0.00806
18950 Pnp 1.82 0.01299
319269 A130040M12Rik 1.81 0.00880
225845 Hrasls3 1.80 0.02078
20128 Trim30 1.80 0.04882
12462 Cct3 1.79 0.04713
73
Table IV. (continued)

Entrez Gene Gene Symbol Ratio p-value
224794 Enpp4 1.79 0.04624
239122 Setdb2 1.79 0.00665
14972 H2-K1 1.78 0.02337
16145 Igtp 1.77 0.03199
246730 Oas1a 1.77 0.00131
17067 Ly6c 1.74 0.04919
21355 Tap2 1.74 0.04987
56048 Lgals8 1.73 0.01375
18712 Pim1 1.72 0.02203
20846 Stat1 1.71 0.02555
16452 Jak2 1.71 0.01782
327959 Fbxo39 1.71 0.00590
219132 D14Ertd668e 1.71 0.00539
27364 Srr 1.70 0.03001
16168 Il15 1.70 0.00584
30935 Tor3a 1.69 0.00494
67877 Nat5 1.69 0.02238
16164 Il13ra1 1.69 0.00029
20773 Sptlc2 1.68 0.03569
224794 Enpp4 1.68 0.04511
211329 Ncoa7 1.67 0.02092
12870 Cp 1.66 0.03997
21859 Timp3 1.66 0.00605
14528 Gch1 1.65 0.01380
14972 H2-K1 1.65 0.02275
19188 Psme2 1.65 0.03988
17698 Msn 1.63 0.00535
16193 Il6 1.63 0.00433
19186 Psme1 1.62 0.02796
15039 H2-T22 1.61 0.00086
21917 Tmpo 1.61 0.00820
80898 Arts1 1.61 0.04648
15950 Ifi203 1.60 0.03540
15953 Ifi47 1.60 0.00218
226517 Smg7 1.60 0.04105
64164 Ifrg15 1.59 0.01949
13163 Daxx 1.59 0.01043
21917 Tmpo 1.58 0.01573
16164 Il13ra1 1.57 0.01763
16950 Loxl3 1.57 0.01813
22169 Tyki 1.56 0.01652
110454 Ly6a 1.55 0.00989
114564 Csprs 1.55 0.04523
231655 Oasl1 1.55 0.04560
74
Table IV. (continued)

Entrez Gene Gene Symbol Ratio p-value
16423 Cd47 1.54 0.01151
16648 Kpna3 1.54 0.02691
54123 Irf7 1.54 0.03333
27050 Rps3 1.54 0.01746
12363 Casp4 1.53 0.01207
12364 Casp12 1.53 0.01247
56736 Rnf14 1.52 0.01610
15959 Ifit3 1.52 0.01306
14701 Gng12 1.52 0.01686
71934 Car13 1.52 0.04470
70110 Ifi35 1.51 0.02413
665317 EG665317 1.51 0.00366
59027 Pbef1 1.50 0.03068
80285 Parp9 1.50 0.04946
70174 2210409B22Rik -1.51 0.01546
56490 Zbtb20 -1.51 0.04391
70164 2210411K19Rik -1.52 0.04392
217410 Trib2 -1.56 0.03889
21858 Timp2 -1.62 0.03598
83397 Akap12 -1.62 0.04212
16997 Ltbp2 -1.64 0.03516
208777 Sned1 -1.66 0.03266
11931 Atp1b1 -1.67 0.01603
21858 Timp2 -1.72 0.04932
18035 Nfkbia -1.72 0.02011
11931 Atp1b1 -1.79 0.00638
20315 Cxcl12 -1.99 0.02292
20315 Cxcl12 -2.05 0.02080
20720 Serpine2 -2.11 0.01667
27528 D0H4S114 -2.32 0.01878

One of the initial observations that we were struck with had to do with the fact that
most of the genes whose expression was affected by Runx2 were up-regulated. This
suggests that Runx2 may behave as more of an activator in this system.
Additionally, we noticed that among the genes whose expression changed the most
in response to Runx2, the chemokines seemed to be over-represented. Specifically,
we saw that Ccl-2, Ccl-5, Ccl-7, and Cxcl-10 were among the top 15 genes that were
75
up-regulated. Moreover, Cxcl-12 was one of the most down-regulated genes.
Chemokines and their receptors are involved in cell migration associated with
development and wound healing; however, they have also been implicated in
osteoclast development. In particular, they help to guide circulating monocytes into
the bone marrow, where they are induced by other signals to differentiate into
osteoclasts. Importantly, these chemokines may be derived from osteoblasts that
inhabit the bone marrow compartment. We hypothesized, therefore, that chemokines
are Runx2 targets, and that although they may not play a direct role in osteoblast
development, they may influence osteoclast development. This is consistent with the
notion that osteoclast and osteoblast development are tightly linked.

To test our hypothesis, we first looked for further evidence that Runx2 does indeed
regulate chemokine expression. To this end, we co-transfected C3H10T1/2 cells
with a Runx2 expression vector and a dominant-negative Runx2 expression
construct. We reasoned that if Runx2 was indeed directly responsible for regulating
chemokine expression, then we should see an inhibition of the effect in the presence
of a dominant-negative version of Runx2. Surprisingly, we did not observe a
statistically significant difference in chemokine expression between cells transfected
with either Runx2 and DN-Runx2 or Runx2 alone (Fig. 17B-G). More disturbing
though, was the fact that we did not even see a statistically significant difference in
chemokine expression between cells transfected with either Runx2 or control vector
(Fig. 17B-G). Indeed, this contradicted previous results, in which we confirmed by
76
RT-qPCR analysis that chemokine expression was stimulated in response to Runx2
(Fig. 17H-L).
77

Fig. 17. RT-qPCR analysis of RNA collected from C3H10T1/2 cells transfected with Runx2.
OC
0
1
2
3
Ccl-2
0
1
2
3
4
5
6
7
8
Cxcl-10
0
1
2
3
4
5
6
7
8
Ccl-7
0
1
2
3
4
Ccl-5
0
1
2
IL-6
0
1
2
3
4
5
6
1 231 23 1 2 312 3
Runx2 DN-R2 Runx2,DN-R2 CMV
IL-6
0
10
20
30
40
50
60
Ccl-2
0
1
2
Ccl-5
0
1
2
3
Ccl-7
0
1
2
3
Cxcl-10
0
1
2
3
4
1 231 23
Runx2 Control
A.
B.
C.
D.
E.
F.
G.
H.
I.
J.
K.
L.
78
To address the cause of this discrepancy, we first considered whether the
experiments had been carried out in exactly the same way. One of the major
differences we could recall was that the Runx2 expression vector used in each
experiment had been derived from different plasmid preps. Thus, we reasoned that
perhaps the effect was due to a difference in plasmid preparation. To test this, we
repeated the transfection on C3H10T1/2 cells with Runx2 expression vector from the
original plasmid prep. Accordingly, we observed that IL-6 expression was
significantly enhanced in response to exogenous Runx2, and this occurred in a dose-
responsive manner (Fig. 18). Although it is not a chemokine, IL-6 expression
behaves similarly, and it may play a role in the differentiation of osteoclasts. Thus,
we concluded that the differences in chemokine and IL-6 expression between the two
experiments were due simply to differences in the quality of plasmid preparations.

Fig. 18. RT-qPCR analysis of IL-6 expression in C3H10T1/2 cells transfected with different
doses of Runx2.

IL-6
0
2
4
6
8
10
12
Runx2

79
Although we knew that one of the plasmid preps was causing experimental artifacts,
we could not tell for certain which one it was. Therefore, we prepared a third
independent Runx2 expression plasmid and transfected it into C3H10T1/2 cells.
Once again, chemokine expression was not significantly altered in the presence of
exogenous Runx2; however, it was up-regulated in response to AML-Eto, a
dominant-negative version of Runx2 (Fig. 19). We concluded from this that the
stimulation of chemokine expression that we observed in the microarray experiment
and subsequent RT-qPCR analysis of the same samples was an experimental artifact.

Fig. 19. RT-qPCR analysis of gene expression in C3H10T1/2 cells transfected with Runx2.

Runx2
0
1
2
Cxcl-10
0
2
4
6
Ccl-7
0
1
2
3
Ccl-5
0
1
2
OC
0
1
2
3
IL-6
0
1
2
3
12 3 123 1 212 12
Runx2 CMV Aml-Eto, Runx2 Aml-Eto, CMV Aml-Eto
Ccl-2
0
1
2
3
A.
B.
C.
D.
E.
F.
G.
80
In spite of the fact that up-regulation of chemokines in the microarray experiment
was not biologically significant, it was not lost on us that OC expression was induced
as well (Fig. 16C, 17A, 19B). Since OC is a known Runx2 target gene, we reasoned
that perhaps the expression of other genuine Runx2 targets was stimulated as well.
Complicating their discovery, of course, was the fact that many genes seemed to be
spuriously up-regulated. One potential explanation for this spurious stimulation of
gene expression is bacterial contamination of the plasmid DNA. Consistent with this
notion, chemokines are likely to be involved in the innate immune response of cells.
Moreover, we observed that many interferon-related genes were up-regulated in the
same experiment, suggesting that the cells were indeed mounting some type of
innate-immune response. We reasoned that if this explanation is correct, then we
should be able to identify genuine Runx2 targets by ignoring those genes that seem
to play a role in innate immunity of the cell. To this end, we re-examined the list of
genes from the microarray experiment and identified several genes whose expression
was stimulated in response to Runx2, but whose biological function seemed
unrelated to innate immunity. We then measured the expression levels of these
genes in C3H10T1/2 cells expressing endogenous Runx2. Contrary to our
hypothesis, none of these genes were stimulated in response to Runx2 (Fig. 20C-F).
Thus, we concluded that further pursuit of Runx2 target genes in this system was not
warranted.
81

Fig. 20. RT-qPCR analysis of gene expression in C3H10T1/2 cells transfected with Runx2.

L10A
0
1
2
OC
0
2
4
6
8
10
Lgals8
0
1
2
3
Ncoa
0
1
2
Setd
0
1
2
3
4
5
Trex
0
1
2
1323 121 2
Runx2 CMV Aml-Eto, Runx2 Aml-Eto, CMV
A.
B.
C.
D.
E.
F.
82
Chapter 6: Optimzation of Critical Parameters for Runx2 Location
Analysis in MC3T3-E1 Pre-Osteoblasts.

In order to characterize the binding pattern of Runx2 in mineralizing osteoblasts by
ChIP-chip, microgram quantities of ChIP DNA are required. This is problematic
because ChIP typically yields only nanogram quantities of DNA. Although pooling
many ChIP samples together is one way of solving this problem, it is not the most
practical solution. Indeed, the most practical way to obtain the microgram quantities
required for ChIP-chip is to amplify the ChIP DNA by one of several methods. The
most commonly-used method is ligation-mediated PCR (LM-PCR). LM-PCR
involves the ligation of universal linkers to the ends of fragmented DNA, which then
permits PCR amplification of the fragments with universal primers. Although we
have used this method to successfully obtain the microgram quantities of ChIP DNA
needed for ChIP-chip, we have found that the quality of this amplified DNA is not
satisfactory for ChIP-chip. Specifically, it is not enriched for Runx2-bound
fragments. We hypothesized that this was because the quantity of starting material
was below a threshold needed to achieve representative amplification.

83

Fig. 21. qPCR analysis of Runx2 ChIP before and after amplification.

To identify the amount of starting material required to achieve representative
amplification, we first performed LM-PCR on serial dilutions of un-enriched input
DNA isolated from MC3T3-E1 cells. Then, we measured by qPCR the copy number
of several genomic loci in the amplified samples as well as in un-amplified samples.
Not surprisingly, we noticed a clear correlation between starting amount of DNA and
success of the amplification. Thus, we were unable even to amplify the smallest
amounts of DNA. Moreover, of the samples that were amplified, all but two
containing the highest amounts of input were not measurable by qPCR. Finally, of
these two samples, the one derived from less input DNA was not representative of
the un-amplified sample, according to qPCR analysis of several genomic loci. Thus,
only the highest amount of input DNA was successfully amplified in a way that
preserved the integrity of the starting material.
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
Runx2 IgG Runx2 IgG
12
Percentage Input
Osteocalcin
Insulin
0
1
2
3
4
5
6
7
Runx2 IgG Runx2 IgG
1st Round 2nd Round
Percentage Input
Insulin
Osteocalcin
A. B.
84

Fig. 22. qPCR analysis of input DNA before and after amplification.

We estimated that the amount of DNA contained in this sample prior to amplification
was about 200ng. According to the limited amount of literature pertaining to
amplification of ChIP material, 10ng of sample should be sufficient for a successful
LM-PCR reaction. Given the disparity between our results and reports from the
literature, we hypothesized that we had overestimated the amounts of input DNA
used in our experiments. This seemed likely, given the opportunity for sample loss
during the clean-up step following RNase treatment. Treatment of the ChIP samples
with RNase prior to amplification is necessary to remove contaminating nucleotides
that might interfere with the reaction. This is typically followed by purification on
columns. We reasoned that a significant portion of the sample was staying on the
columns, and that the amount of sample getting amplified was less than our original
estimate. Thus, according to our hypothesis, we were still working with sub-optimal
amounts of DNA for LM-PCR. To test this hypothesis, we repeated the
amplification with 3 different amounts of input DNA that had been RNase treated
16
21
26
31
36
200 ng 100 ng
Pre-LMPCR Post-LMPCR
Mean Ct Value
OPN (-240/-438)
OPN (-39/-263)
OG2 (-86/-285)
OG2 (-6/-196)
Myo (-172/-393)
Myo (-19/-267)
Col (-1310/-1470)
Bsp (-14/-246)
Bsp(-1322/-1499)
85
and either column-purified, or purified by phenol/chloroform extraction, followed by
ethanol precipitation. If our hypothesis was correct, then the ethanol-precipitated
samples should amplify better than the column-purified samples. As expected, more
DNA was recovered by ethanol precipitation than by column purification; however,
it was the column-purified samples that gave the greatest yields after LM-PCR.
Moreover, when the amplified samples were analyzed by real-time PCR, ratios
between almost all the loci measured were significantly different from those of the
un-amplified material. This was true for both the column-purified and the ethanol-
purified DNA.
86

Fig. 23. qPCR analysis of input DNA before and after amplification.

Coincident with these findings, a paper from the Farnham lab was published that
described similar problems with LM-PCR, as well a potential solution. According to
the paper, ChIP DNA that was amplified by the commonly-used LM-PCR protocol
compared poorly to un-amplified ChIP DNA in ChIP-chip experiments. Although
the underlying reason for this was not given, it may be due to the difficulty inherent
in achieving blunted ends for ligation (Peggy Farnham, personal communication).
Regardless, the paper advocated the use of a whole genome amplification kit
available commercially from Sigma for the purpose of amplifying ChIP DNA for
ChIP-chip. Wanting to take advantage of this timely information, we used the kit as
a platform on which to repeat the previous experiment, where 3 different amounts of
15
20
25
30
35
40
Column EtoH Column EtoH Column EtoH
Pre-LMPCR Post-LMPCR Pre-LMPCR Post-LMPCR Pre-LMPCR Post-LMPCR
1 ng 10 ng 100 ng
Mean Ct Value
Bsp (-14/-246) Bsp (-1322/-1499) Ins (+37/-161) Og2 (-86/-285) Opn (-240/-438)
87
DNA were amplified. To our disappointment, we obtained similar results, in which
the amplified samples were not representative of the input DNA.

Fig. 24. qPCR analysis of input DNA before and after amplification.

A potential explanation for the failure of this experiment became apparent when we
investigated the details of the whole genome amplification kit. The kit was designed
to amplify genomic DNA by converting it to PCR-amplifiable molecules flanked by
universal priming sites and by PCR amplification with universal oligonucleotide
primers and a limited number of cycles. Conversion to PCR-amplifiable molecules
involves a chemical fragmentation step that breaks the DNA up into small pieces.
Since ChIP DNA is already fragmented, the fragmentation step is unnecessary, and
therefore can be omitted from the WGA protocol. Failure to add the fragmentation
20
22
24
26
28
30
32
34
36
38
40
Column EtOH Column EtOH Column EtOH
-Pre
LMPCR
Post-LMPCR -Pre
LMPCR
Post-LMPCR -Pre
LMPCR
Post-LMPCR
1ng Input DNA 10ng Input DNA 100ng Input DNA
Mean Ct Value
Bsp (-14/-246) Bsp (-1322/-1499) Opn (-240/-438) OG2 (-86/-285)
88
buffer may adversely affect downstream portions of the protocol though, and so we
repeated the experiment in the presence of fragmentation buffer without permitting
the fragmentation to actually take place. Surprisingly, only one of the two input
samples was amplified, and it showed a significant amount of bias upon analysis by
real-time PCR. Thus, we concluded that some intrinsic property of our ChIP DNA
was responsible for the amplification bias.

Fig. 25. qPCR analysis of input DNA before and after amplification.

To address the possibility that the ChIP procedure was somehow making the DNA
unsuitable for representative amplification, we tested several hypotheses in parallel.
Specifically, we tested to see whether the cross-linking, sonication, or reversal of
cross-links could be the culpable steps. To this end, we amplified 10 ng of the
following samples: 1) DNA that had been cross-linked, sonicated for various lengths
of time, and then undergone cross-link reversal; 2) DNA that had not been cross-
linked, but had been sonicated for various lengths of time and had undergone
crosslink reversal; 3) DNA that had not been cross-linked, but had been sonicated
15
20
25
30
35
40
15ng 150ng 15ng 150ng
Pre-LMPCR Post-LMPCR
Mean Ct Value
Bsp (-14/-246) Bsp (-1322/-1499) OG2 (-86/-285) OG2 (-6/-196)
Myo (-172/-393) Ins (-10/-233) Ins (+37/-161) Opn (-240/-438)
89
and did not undergo cross-link reversal; 4) NIH 3T3 DNA available from a
commercial source. Unexpectedly, all of these samples showed bias as a result of
amplification. That this was true, even for the genomic DNA sample led us to
believe that the bias in amplification was not due to any intrinsic property of our
ChIP DNA, but rather was an artifact introduced by the kit. Thus, amplification bias
seems to be an unavoidable problem that plagues high-througput location analysis
techniques such as ChIP-chip.

Fig. 26. qPCR analysis of input DNA before and after amplification.

In spite of the fact that amplification bias appears to be unavoidable, meaningful
results can still be obtained by location analysis, provided that one is careful to limit
the bias as much as possible. In view of this, two discoveries were made during our
troubleshooting attempts that may help to limit amplification bias. First,
formaldehyde cross-linking seemed to greatly affect the isolation and subsequent
fragmentation of chromatin from MC3T3-E1 cells. Second, no matter what the size
21
22
23
24
25
26
27
28
Pre Post Pre Post Pre Post Pre Post Pre Post Pre Post
No Sonication Short Sonication Long Sonication Short Sonication Long Sonication
Uncrosslinked Crosslinked
NIH 3T3 DNA MC3T3 DNA
Mean Ct Value
OG2 (-86/-285) Ins (+37/-161) Bsp (-14/-246) Opn (-240/-438)
90
range of input material was, the output size range was always between 1000 and 200
base pairs, with the average size being about 500 base pairs. Thus, it appeared as
though only smaller molecular weight fragments were being amplified. The
significance of this is that the majority of input DNA from cross-linked cells was not
being amplified, since it was high molecular weight. Because the amount of bias
correlates well with the amount of starting material, insufficient chromatin
fragmentation may indirectly introduce a substantial amount of amplification bias.

Fig. 27. Gel analysis of sonicated MC3T3-E1 chromatin before and after amplification.

To address the issue of chromatin fragmentation, we first tested the hypothesis that
poor fragmentation was the result of chromatin being cross-linked to the heavy
extracellular matrix secreted by these cells. To this end, we isolated chromatin from
C4-2B cells, as well as early stage MC3T3-E1 cells which do not have much
extracellular matrix. The efficiency of isolation was equally poor after cross-linking
for these cells as it was for the late-stage MC3T3-E1 cells, thereby disproving our
01 23 4 5 01 2 3 4 5 0 1 2 3 4 5 Pulses:
X-link No X-link
1600 bp
500 bp
01 23 4 5 01 2 3 4 5 0 1 2 3 4 5 Pulses:
X-link No X-link
1600 bp
500 bp
No X-link
02 4 Pulses:
X-link
02 4 0 2 4
3T3
DNA
1600 bp
500 bp
No X-link
02 4 Pulses:
X-link
02 4 0 2 4
3T3
DNA
1600 bp
500 bp
A. B.
91
hypothesis. We next tested the hypothesis that we weren’t sonicating the cross-
linked chromatin sufficiently. To this end, we sonicated at higher settings. When
this proved largely ineffective, we used a more powerful sonicator. Initially, we
were encouraged to see a low-molecular weight smear from both cross-linked and
non-cross-linked cells; however, these results were not reproducible.
92

Fig. 28. Gel analysis of sonicated MC3T3-E1 chromatin.

24 6 Pulses:
No
C4-2B
1600 bp
500 bp
24 62 4 6 24 6 2 4 6 24 6
X-link No X-link No X-link
MC3T3-E1 (early) MC3T3-E1 (late)
24 6 Pulses:
No
C4-2B
1600 bp
500 bp
24 62 4 6 24 6 2 4 6 24 6
X-link No X-link No X-link
MC3T3-E1 (early) MC3T3-E1 (late)
12 3 Pulses:
1600 bp
500 bp
45 6
X-link No
12 3 4 5 6 12 3 Pulses:
1600 bp
500 bp
45 6
X-link No
12 3 4 5 6
12 3 Pulses:
1600 bp
500 bp
45 6 1 2 3 45 6 1 2 3 45 6
4
812
Power Setting:
12 3 Pulses:
1600 bp
500 bp
45 6 1 2 3 45 6 1 2 3 45 6
4
812
Power Setting:
A.
B.
C.
93
A potential explanation for the non-reproducibility of our results had to do with a
slight deviation from the normal protocol that occurred during the one “successful”
experiment with a more powerful sonicator. Specifically, the formaldehyde cross-
links had been reversed by boiling for 15 minutes, rather than a 65°C incubation
overnight. To see if this deviation could account for the low-molecular weight smear
that we presumed to be fragmented chromatin, we reversed the cross-links of
sonicated chromatin by the following incubations: 1) 65°C overnight, 2) 65°C for 4
hours, 3) 75°C for 4 hours, 4) 85°C for 4 hours, and 5) boiling for 15 minutes. The
results demonstrated that boiling for 15 minutes was indeed responsible for the low-
molecular weight smear. Although we hoped that the problem of fragment size
could be solved simply by altering our protocol for reversing the cross-links, the
possibility remained that the low molecular weight species on the gel was RNA
instead of DNA. To test this, we isolated cross-linked chromatin, reversed the cross-
links by boiling for various lengths of time, and then treated half the samples with
RNase. This experiment proved that the low molecular weight smear was indeed
RNA.

94

Fig. 29. Gel analysis of sonicated MC3T3-E1 chromatin.

Given our inability to reduce chromatin fragment size by increasing the sonication
intensity, our next approach was to reduce the amount of cross-linking. Thus, we
cross-linked MC3T3-E1 cells for various lengths of time and then isolated the
chromatin. After reversing the cross-links, purifying the DNA, and visualizing it by
gel electrophoresis, we saw fragments of the appropriate size range after 2.5 minutes
of cross-linking and several pulses of sonication.

A. B.
65°
2.0 kb
600 bp
O/N 4hr 4hr4hr15min
75° 85° 100°
65°
2.0 kb
600 bp
O/N 4hr 4hr4hr15min
75° 85° 100°
7.5’
2.0 kb
600 bp
+-+-+-+-+-+ -
15’ 30’ 7.5’ 15’ 30’
5 10
RNase:
7.5’
2.0 kb
600 bp
+-+-+-+-+-+ -
15’ 30’ 7.5’ 15’ 30’
5 10
RNase:
95

Fig. 30. Gel analysis of sonicated MC3T3-E1 chromatin.

To test whether or not the abbreviated cross-linking time was sufficient to facilitate a
successful immunoprecipitation, we performed Runx2 ChIP on this chromatin,
followed by qPCR analysis. Absence of enrichment led us to conclude that
shortened cross-linking times may not work well for ChIP. Two possibilities may
explain this. First, the experiment may have failed for technical reasons, apart from
the cross-linking. Consistent with this possibility, even the standard 10-minute
cross-linking time resulted in an unsuccessful ChIP. Second, the experiment may
have failed because we destroyed the epitope recognized by the Runx2 antibody
during our more aggressive sonication procedure. This possibility means that a
delicate balance must be struck between over-sonicating, which would result in
epitope loss, and under-sonicating, which would result in chromatin fragments of the
wrong molecular weight.

3
2.0 kb
600 bp
0’
69 123
2.5’
69 12 3
5’
69 12 3
10’
69 12 3
2.0 kb
600 bp
0’
69 123
2.5’
69 12 3
5’
69 12 3
10’
69 12 4
2.5’
812 4
5’
8124
10’
812
1600 bp
500 bp
4
2.5’
812 4
5’
8124
10’
812
1600 bp
500 bp
A. B.
96

Fig. 31. qPCR analysis of Runx2 ChIP performed under various conditions.

To address the first possibility, we repeated the experiment. Interestingly, the length
of cross-linking time did not appear to dramatically affect the total amount of DNA
that was recovered. Moreover, sonication strength did not appear to dramatically
improve the yield of DNA recovery. This was true regardless of cross-linking
length. Finally, there is no clear correlation between the number of sonication pulses
and the amount or size of DNA that was recovered. In spite of these unexpected
results, we decided to immunoprecipitate some of the samples. Real-time analysis of
the ChIP samples shows that there was specific enrichment in every case.
Interestingly, 10 minutes of cross-linking appeared to result in greater overall
enrichment compared to 2.5 minutes. Among the samples that were derived from
cells that had been cross-linked for 2.5 minutes, there appeared to be a trend towards
greater enrichment with increasing sonication strength and time. This did not appear
to be the case for the samples derived from cells that had been cross-linked for 10
minutes. Although non-specific enrichment was minimal or absent from all the
samples, the overall levels of the regions analyzed (OC and Ins) were abnormally
high in IgG control samples obtained from chromatin isolated by the standard ChIP
A. B.
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
4 pulses 8 pulses 4 pulses 8 pulses 4 pulses 8 pulses
2.5 min 5 min 10 min
Percentage Input
Osteocalcin
Insulin
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
8 pulses 12 pulses 8 pulses 12 pulses 8 pulses 12 pulses
2.5 min 5 min 10 min
Percentage Input
Osteocalcin
Insulin
97
246 24 624 62 46 2462 46 # Pulses:
Sonication
Strength:
4W 8W 16W 4W 8W 16W
Cross-linking
Time: 2.5 minutes 10 minutes
1.6 Kb
0.5 Kb
0
0.5
1
1.5
2
2.5
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
4 pulses 6 pulses 4 pulses 4 pulses 4 pulses 6 pulses
4 Watts 8 Watts 16 Watts 4 Watts 8 Watts
2.5 min 10 min
Percentage Input
Osteocalcin
Insulin
protocol (10-minutes cross-linking, 4 Watts, 4 pulses). The significance of this is
unclear; however, it is possible that it is merely a chance occurrence.
Fig. 32. Gel and ChIP analysis of cross-linked chromatin isolated from MC3T3-E1 cells.

To ascertain whether the results were reproducible, we repeated it with a subset of
sonication and cross-linking times. We found that the samples were fairly
homogenous with respect to fragment size, especially those derived from cells that
were cross-linked for 10 minutes. Moreover, the average fragment size appeared to
be between 2 and 3 Kb. Finally, there did not appear to be more DNA in the samples
from cells that were cross-linked for 10 minutes than in those from cells that were
cross-linked for 2.5 minutes. With respect to enrichment, there was not a dramatic
difference between the samples derived from cells cross-linked for 10 minutes and
those derived from cells cross-linked for 2.5 minutes. In both groups there was
enrichment for the OSE2 element of the osteocalcin promoter in the samples
immunoprecipitated with Runx2 antibody as well as in the samples
immuniprecipitated with IgG. The specific enrichment appeared to be slightly
A.
B.
98
12 31 2 3
2.5’
10’
1.6 Kb
0.5 Kb
Cross-linking Time:
Sample #:
2.0 Kb
3.0 Kb
0
0.5
1
1.5
2
2.5
3
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
#1 #2 #3 #1 #2 #3
2.5 min 10 min
Percentage Input
Osteocalcin
Insulin
greater for the group of samples derived from cells that had been cross-linked for 10
minutes, while the non-specific enrichment appeared to be slightly greater for the
group derived from cells that had been cross-linked for 2.5 minutes. Thus, shorter
cross-linking times not only failed to give more input DNA, they also failed to
provide enrichment levels that are comparable to what can be obtained with standard
conditions.

Fig. 33. Gel and ChIP analysis of cross-linked chromatin isolated from MC3T3-E1 cells.

At this point, we recalled that 5 minutes of cross-linking combined with extensive
sonication (twelve 5-second pulses) gave fragments that were close to the
appropriate size in two instances. To ascertain whether 5 minutes of cross-linking is
sufficient to yield acceptable enrichment by ChIP, twelve 100-mm plates of day 8
MC3T3-E1 cells were cross-linked for 5 minutes with 1% formaldehyde, collected
into 6 tubes (2 plates/tube), and sonicated for a various number of 5-second pulses.
Only three tubes survived 12 such pulses (~1 minute) without foaming. The
A.
B.
99
remaining three tubes were centrifuged for 10 minutes at 13,000rpm (4°C) and then
subjected to a various number of longer pulses (~15 seconds) on ice. Interestingly,
the chromatin solution in these latter samples was significantly clarified by the
lengthened sonication. After a second centrifugation, most of them consisted of a
clear, yellowish solution and a small white pellet in the bottom of the tube. In
contrast, the three samples that had been sonicated for only 1 minute consisted
mostly of an opaque, yellowish solution. A small white pellet could be made out in
the bottom of the tube, and there was a small amount of clear, yellowish solution at
the top. Approximately 200ul was taken from each sample, using care not to disturb
the pellet at the bottom. Of this, 100ul was immunoprecipitated with Runx2
antibody, and 100ul was immunoprecipitated with non-specific antibody (IgG).
Additionally, 10ul was taken from each sample as input DNA. The cross-links of the
input DNA were reversed overnight at 65°C, followed by protein and RNA removal.
One-third of the resulting DNA was electrophoresed on 1% agarose stained with
EtBr. With 5 minutes of cross-linking and 1 minute of sonication, the average
chromatin fragment size was about 500bp. Interestingly, with longer sonication
times, the average fragment size became even smaller. After cross-link reversal,
protein removal, and DNA extraction, enrichment was measured in the 12 ChIP
samples by real-time PCR. Although there was a clear trend towards enrichment in
all of the Runx2-ChIP samples, this same trend appeared in many of the
accompanying IgG-ChIP samples as well. For the samples derived from chromatin
that was sonicated for more than 1 minute, the amount of enrichment achieved by IP
100
1.6
0.5
2.0
3.0
MW
(Kb)
0.4
0.3
0.2
12 3
3’ 2’ 1’
Time of
Sonication : 4’
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
#1 #2 #3
1 minute 2 minutes 3 minutes 4 minutes
Percentage Input
Osteocalcin
Insulin
with IgG was comparable (if not greater) than that which was achieved with Runx2-
specific antibody. For the samples derived from chromatin that was sonicated for 1
minute, the amount of enrichment achieved by IP with IgG was consistently less than
that which was achieved with Runx2-specific antibody. Still, the amount of specific
enrichment was less than what can be achieved with standard ChIP conditions, while
the opposite was true of the non-specific enrichment.

Fig. 34. Gel and ChIP analysis of cross-linked chromatin isolated from MC3T3-E1 cells.

One possible interpretation of these results is that shorter cross-linking and longer
sonication times are sub-optimal for ChIP. That is, the samples which were
sonicated for only 1 minute showed poor enrichment levels because they were not
cross-linked sufficiently, and the samples that were sonicated for longer periods of
time showed even poorer enrichment because the epitope for antibody recognition
has been destroyed. Of course, no positive controls were included to provide
assurance that this experiment did not merely fail for technical reasons. Assuming it
A.
B.
101
did not though, and that our interpretations were valid, we reasoned that it would
make sense to repeat this experiment with longer cross-linking, shorter sonication
times, and a positive control.

To this end, the experiment was repeated on chromatin from day 8 MC3T3-E1 cells
that had been cross-linked for 10 minutes. These samples were sonicated at medium
power (~5-6 Watts) for 15-second pulses on ice. Sample 1 was sonicated for one
pulse, sample 2 for 2 pulses, etc. All the samples were then centrifuged for 10
minutes at 13000 rpm (4°C). Samples 5 and 6 were clear except for a small white
pellet in the bottom of the tube. The other samples remained cloudy and viscous. A
200-ul aliquot was taken from each tube, half of which was immunoprecipitated with
Runx2 antibody and half of which was immunoprecipitated with IgG. A 10-ul
aliquot was kept as input DNA.After cross-link reversal, protein removal, and DNA
extraction, enrichment was measured in the 12 ChIP samples by real-time PCR.
Overall, it appeared as though this ChIP was much more successful than the previous
one. There was a significant amount of enrichment in each of the Runx2 ChIP
samples that was not seen (at least not to the same extent) in the IgG samples. While
the experiment was a success in terms of enrichment, the average fragment size for
each of these samples was well above the ideal 500-bp range we had hoped to
achieve.

102
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
15'' 30'' 45'' 60'' 72'' 96''
Percentage Input
Osteocalcin
Insulin
1.6
2.0
3.0
MW
(K b )
4.0
5.0
6.0
45’’ 15’’ 60’’ 72’’ 96’’

Fig. 35. Gel and ChIP analysis of cross-linked chromatin isolated from MC3T3-E1 cells.

Since it appears that 96 seconds of sonication did not negatively affect the amount of
enrichment we were able to achieve, it seemed reasonable to repeat the previous
experiment with longer sonication times, making sure as before to include a positive
control. Thus, we plated twelve 100-mm dishes and one 12-well plate of MC3T3-E1
cells, then treated them (or not) with 50 ug/ul ascorbic acid for 8 days. On the eighth
day (today), we cross-linked the cells for 10 minutes each, then sonicated each of the
6 samples at ~4 Watts for various numbers of 15-second pulses on ice. One sample
was sonicated for only a single pulse, and served as my positive control. The other 5
samples received 2, 3, 4, 5, or 6 minutes (total) of sonication. After centrifugation,
all the samples (including the control) were clear (and yellowish) with a small white
pellet in the bottom of the tube. Two-hundred microliters of this solution was taken
from each sample for immunoprecipitation (100ul per IP: Runx2 and IgG), and 10ul
was taken as input. Again, there appeared to be enrichment for osteocalcin, but this
was seen in both the Runx2 and IgG samples, especially for those derived from cells
that were sonicated longer than 2 minutes.Regardless, gel electrophoresis of the input
A. B.
103
DNA (before and after RNase treatment) showed that the average fragment size of
all the samples was well above 500bp.

Fig. 36. Gel and ChIP analysis of cross-linked chromatin isolated from MC3T3-E1 cells.

It seems as though chromatin from cells that have been cross-linked for 10 minutes is
extremely resistant to shearing by sonication. Although it is possible that longer
C.
1.6
2.0
3.0
MW
(Kb)
4.0
5.0
6.0
0.5
0.4
0.3
0.2
2’ 15’’ 3’ 4’ 5’ 6’
Time of
Sonication :
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
15'' 2'3' 4'5' 6'
Percentage Input
Osteocalcin
Insulin
Before RNase After RNase
A.
B.
1.6
2.0
3.0
MW
(Kb)
4.0
5.0
0.5
0.4
0.3
0.2
2’ 15’’ 3’ 4’ 5’ 6’
Time of
Sonication :
104
sonication periods may yield the desired fragment size, 6-minutes already is an
uncomfortably long period of time. Given this, we re-considered cross-linking times
of less than 10 minutes. Although we knew that 5 minutes was too short to achieve
acceptable enrichment, we reasoned that there might be a time between 5 and 10
minutes that is optimal for both fragment size and enrichment. Thus, we plated
twelve 100-mm dishes of MC3T3-E1 cells and cultured them in the presence of
ascorbic acid for 8 days. They were then cross-linked for various lengths of time
between 5 and 10 minutes. There were 12 plates, and they were processed in pairs.
One pair saw 5 minutes of 1% formaldehyde before collection, another pair saw 6
minutes, another 7, and so on. After collection, the 6 samples were all sonicated in
lysis buffer for 1 minute (15-second pulses) on ice. They were centrifuged, and a 10-
ul aliquot was taken from the supernatant of each sample. Interestingly, the white
pellet at the bottom was largest in the samples derived from cells that had been cross-
linked for the shortest period of time. The cross-links of the aliquots were reversed
overnight at 65°C, then digested with Proteinase K, phenol/chloroform extracted,
ethanol precipitated, and RNase digested before electrophoresis on a 1% agarose gel
stained with EtBr. Interestingly, all six samples looked very similar, with the
average fragment size concentrated around 4 Kb. Thus, the remainder of each
sample was thawed and sonicated for an additional minute (15-second pulses) on ice.
After centrifugation, the pellets at the bottom of the tubes looked much more similar
in size than they did yesterday, with the exception of the sample derived from cells
that had been cross-linked for 9 minutes. The pellet in this sample was considerably
105
larger. Again, a 10-ul aliquot was removed from each sample and the cross-links
were reversed overnight at 65°C, followed by digestion with Proteinase K,
phenol/chloroform extraction, ethanol precipitation, and RNase digestion before
electrophoresis on a 1% agarose gel stained with EtBr. Once again, the average
fragment size was well above 500 bp. Interestingly, the intensity of several samples
had diminished somewhat compared to the previous analysis. Thus, the remainder of
the samples was thawed and sonicated for an additional minute (15-second pulses)
on ice. After centrifugation, the pellets at the bottom of the tubes all looked similar
in size, including the sample derived from cells that had been cross-linked for 9
minutes. Again, a 10-ul aliquot was removed from each sample and the cross-links
were reversed overnight at 65°C, followed by digestion with Proteinase K,
phenol/chloroform extraction, ethanol precipitation, and RNase digestion before
electrophoresis on a 1% agarose gel stained with EtBr. Interestingly, there was a
subtle shift in the size distribution of the fragments. Specifically, it appeared as
though the average size of the fragments was closer to 500bp, particularly in the
samples derived from cells that had been cross-linked for 7 minutes or less. Thus,
we proceeded with ChIP analysis of these samples.

106

Fig. 37. Gel analysis of chromatin isolated from MC3T3-E1 cells.

After overnight immunoprecipitation with either Runx2-specific antibody or IgG, the
unbound DNA was washed away, and the crosslinks were reversed overnight at
65°C. After digestion with proteinase K, extraction with phenol/chloroform, and
precipitation with ethanol, they were then analyzed by qPCR. Once again, there was
enrichment for osteocalcin in all the samples measured, including the IgG-
immunoprecipitated samples. On average though, the amount of enrichment seemed
to be greater for the Runx2-immunoprecipitated samples. There also seemed to be a
trend towards greater enrichment with longer cross-linking, both in the Runx2- and
IgG-immunoprecipitated samples.
A. B.
1.6
2.0
3.0
MW
(K b )
4.0
5.0
0.5
0.4
0.3
6’ 7’ 5’ 9’ 10’ 8’
Tim e of
C ross -lin king :
1.6
2.0
3.0
MW
(K b )
4.0
5.0
0.5
0.4
0.3
6’ 7’ 5’ 9’ 10’ 8’
Tim e o f
C ros s -lin k in g :
1.6
2.0
3.0
MW
(K b)
4.0
5.0
0.5
0.4
0.3
6’ 7’ 5’ 9’ 10’ 8’
Tim e of
C ross-linking :
C.
107
Fig. 38. ChIP analysis of chromatin isolated from MC3T3-E1 cells.

Since the chromatin in the last experiment was so resistant to shearing, even after
only 5 minutes of cross-linking, we next pursued a wider range of cross-linking
times (i.e., 3-10 minutes), as well as quenching the cross-linking reaction with
125mM glycine. The range of cross-linking times was between 4 and 10 minutes,
with two samples having been cross-linked for 5 minutes. After collection, the
samples were sonicated on ice for 1 minute each (15-second pulses). Our rationale
for doing so was based on the inclusion of shorter cross-linking times (4-minutes)
and quenching of the reaction with glycine. In short, we did not know how
dramatically these two variables (especially the latter) would affect the resistance of
the chromatin to shearing. Not wishing to over-sonicate the samples, we elected to
proceed conservatively. After sonication, we reversed the cross-links of a 10-ul
aliquot and snap froze the remainder for storage at -80°C. After reversal of the
cross-links, the samples were proteinase K digested, phenol/chloroform extracted,
ethanol precipitated, and RNase-treated. The purified samples were then
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
5' 6' 7' 8' 9' 10'
Percentage Input
Osteocalcin
Insulin
108
electrophoresed on 1% agarose stained with EtBr. The average molecular weight of
the fragments in all 8 samples was well above 500bp. Therefore, we thawed out the
remainder of the samples and sonicated them each for an additional 2 minutes (30-
second pulses) on ice. Our rationale for sonicating an extra 2 minutes (instead of just
one) was based on the observation that it took 3 minutes of sonication to achieve
acceptable fragment sizes in the previous experiment. After sonication, we reversed
the cross-links of a 10-ul aliquot and snap-froze the remainder for storage at -80°C.
After reversal of the cross-links, the samples were proteinase K-digested,
phenol/chloroform-extracted, ethanol-precipitated, and RNase-treated. The purified
samples were then electrophoresed on 1% agarose stained with EtBr. This time, the
average molecular weight of the fragments was much closer to 500bp, in agreement
with the previous experiment. Thus, we proceeded to immunoprecipitate 6 of the 8
samples, excluding the two with the highest molecular weight fragments (the first of
the two 5-minutes cross-linked samples and the 7-minutes cross-linked sample).
109

Fig. 39. Gel analysis of chromatin isolated from MC3T3-E1 cells.

After overnight immunoprecipitation with either Runx2-specific antibody or IgG, the
unbound DNA was washed away, and the crosslinks were reversed overnight at
65°C. Then, the samples were digested with Proteinase K, phenol/chloroform-
extracted, ethanol-precipitated, and analyzed by real-time PCR. Once again, there
was consistent enrichment for the OSE2 element of the osteocalcin promoter. This
was true for all the samples analyzed, including those that were immunoprecipitated
with IgG. Overall though, there did appear to be a greater amount of enrichment in
the samples immunoprecipitated with Runx2-specific antibody. Still, the level of
enrichment was not more than 5-fold at best, and there was no readily observable
correlation between cross-linking time and enrichment amount.
A. B.
1.6
2.0
3.0
MW
(Kb)
4.0
5.0
0.5
0.4
0.3
5’ 5’ 4’ 7’ 8’ 6’
Time of
Cross-linking :
9’ 10’
1.6
2.0
3.0
MW
(Kb)
4.0
5.0
0.5
0.4
0.3
5’ 5’ 4’ 7’ 8’ 6’
Time of
Cross-linking :
9’ 10’
110
Fig. 40. ChIP analysis of chromatin isolated from MC3T3-E1 cells.

Although the amount of enrichment we achieved in the last experiment was sub-
optimal, we reasoned that this could be improved by collecting cells at a later time
point. The rationale behind this is as follows. First, there is a significant increase in
osteocalcin expression around day 14 in MC3T3-E1 cells. If there is a direct
correlation between Runx2 binding (to the OSE2 element of the OC promoter) and
osteocalcin expression, then by performing ChIP on (or around) day 14, greater
enrichment may be achieved than on day 8, which is when ChIP is usually
performed. To this end, we plated twelve 100-mm dishes and cultured them in the
presence of ascorbic acid for 14 days. We cross-linked these cells for either 2.5, 5,
or 10 minutes and quenched the reaction s with 125mM glycine. After collection,
we sonicated each of the samples on ice for 3 minutes using 30-second pulses. We
then removed 10-ul aliquots and snap-froze the remainder for storage at -80°C. The
aliquots were then incubated overnight at 65°C to reverse the cross-links. After
reversal of the cross-links, the samples were proteinase K-digested,
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
4' 5' 6' 8' 9' 10'
Percentage Input
Osteocalcin
Insulin
111
phenol/chloroform-extracted, ethanol-precipitated, and RNase-treated. The purified
samples were then electrophoresed on 1% agarose stained with EtBr. The average
fragment size of the samples was higher than 500bp, with the possible exception of
the samples derived from cells that were cross-linked for only 2.5 minutes. Since
there were two samples from each group of cross-linking times, we sonicated one
sample from each group for an additional 3 minutes on ice (30-second pulses), for a
total of 6 minutes. After this, we immunoprecipitated all the samples in parallel with
either Runx2-specific antibody or IgG. In addition, we saved a 10-ul aliquot from
each sample for fragment size analysis. After incubation overnight at 65°C to reverse
the cross-links, the aliquots were proteinase K-digested, phenol/chloroform-
extracted, ethanol-precipitated, and RNase-treated. They were then electrophoresed
on 1% agarose stained with EtBr. Visualization with UV light revealed that the
average fragment size of the samples that had been sonicated for 6 minutes was
significantly less than that of the samples which had only been sonicated for 3
minutes. Indeed, the average fragment size of these samples was close to the desired
size of 500bp. Interestingly, there was a dark smear at the bottom of the lanes
containing these samples, arousing suspicion of RNA contamination; however,
further treatment with RNase and re-analysis revealed that this was not the case (data
not shown).

112

Fig. 41. Gel analysis of chromatin isolated from MC3T3-E1 cells.

Finally, unbound DNA from the immunoprecipitation was washed away, and the
crosslinks were reversed overnight at 65°C. After cross-link reversal, the samples
were digested with Proteinase K, extracted with phenol/chloroform, precipitated with
ethanol, and analyzed by real-time PCR. There was again enrichment for the OSE2
element of the osteocalcin promoter in nearly all the samples analyzed, although
some could not be analyzed due to low signal levels. The amount of enrichment in
the IgG samples was limited to 2-fold or less, as before. The enrichment in the
Runx2 samples was much greater this time though. In at least one sample from each
of the three groups, the enrichment was 10-fold or more. In one sample (5 minutes
cross-linked, 6 minutes sonicated), the enrichment was nearly 20-fold. Interestingly,
there did not appear to be a strong correlation between the amount of sonication
and/or cross-linking and the amount of enrichment. Significantly, the sample that
B.
1.6
2.0
3.0
MW
(Kb)
4.0
5.0
0.5
0.4
0.3
2.5’ 10’ 5’
Sonication:
Cross-linking :
3’ 6’ 3’ 6’ 3’ 6’
1.6
2.0
3.0
MW
(Kb)
4.0
5.0
0.5
0.4
0.3
2.5’ 10’ 5’ Time of Cross-linking :
A.
113
had the most enrichment was derived from chromatin that had been sonicated for 6
minutes, thus making it a perfect candidate for ChIP-chip.
Fig. 42. ChIP analysis of chromatin isolated from MC3T3-E1 cells.

In spite of the fact that the sample which had been cross-linked for 5 minutes and
sonicated for 6 minutes appeared to be ideal for ChIP-chip, we ended up using the
sample that had been cross-linked for 10 minutes and sonicated for 3 minutes. After
several rounds of amplification, the enrichment that was present in this sample had
not diminished (Fig. 41A), and the fragments were within the appropriate size range
(Fig. 41B).
0
0.5
1
1.5
2
2.5
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
Runx2
IgG
3' 6' 3' 6' 3' 6'
2.5' 5' 10'
Percentage Input
Osteocalcin
Insulin
114

Fig. 43. Gel and qPCR analysis of amplified ChIP material.

0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
Bsp (-1322/-1499)
Bsp (-14/-246)
Ins (+31/-161)
Myo (-172/-393)
OG2 (-86/-285)
OPN (-240/-438)
RANKL (-438/-230)
RANKL (-317/-81)
ChIP
DNA
Input
DNA
MW
Marker
A. B.
115

Fig. 44. Gel and qPCR analysis of amplified ChIP material.
0
0.2
0.4
0.6
OC
Ins
0
0.5
1
1.5
2
ChIP
DNA
Input
DNA
MW
Marker
500 bp -
1600 bp -
A. B.
C.
116
Chapter 7: Progressive Recruitment of Runx2 to Genomic Targets
despite Decreasing Expression during Osteoblast Differentiation

5.1 Introduction
Osteocalcin (OC) is a γ-carboxylated protein produced primarily by osteoblasts
during bone formation. It is both incorporated into the bone extracellular matrix
(ECM) and secreted into the circulation, where for decades it has been used
clinically as a marker of bone turnover [Lian and Gundberg, 1988]. Additionally,
circulating non-carboxylated OC may play a central role in energy metabolism [Lee
et al., 2007]. OC expression is first detectable on embryonic day 15.5, concomitant
with mineralization [Desbois et al., 1994]. Likewise, in several osteoblast culture
models, OC mRNA levels and promoter activity have been shown to peak late
during development of the bone phenotype [Frenkel et al., 1997; Frenkel et al., 1996;
Owen et al., 1990; Quarles et al., 1997; Xiao et al., 1997]. Despite these striking
associations, OC is not necessary for bone formation or development [Ducy et al.,
1996]. Still, tissue-restricted developmentally-controlled transcription from the OC
promoter has rendered it a popular research tool in the pursuit of transcriptional
regulatory mechanisms of osteoblast differentiation and bone formation (Lian, 1998,
J Cell Biochem). Most notably, the master osteoblast transcription factor, Runx2,
was discovered by way of its interaction with the so-called Osteoblast-Specific
Element 2 (OSE2) of the OC promoter [Banerjee et al., 1997; Ducy and Karsenty,
1995; Ducy et al., 1997; Merriman et al., 1995]. Despite the pivotal role of Runx2 in
117
OC transcriptional control, we report here that its expression pattern cannot explain
OC’s developmental expression pattern. In fact, we show that Runx2 expression is
anti-parallel to that of OC, and that the developmental up-regulation of OC
transcription is associated with increased accessibility to Runx2.

5.2 Results
Paradoxical inverse relationship between Runx2 and osteocalcin expression during
osteoblast differentiation.
OC mRNA levels steadily increase as the osteoblast phenotype develops in various
models, including MC3T3-E1 cultures (Fig. 11A). Because OC is a well-established
target of the osteoblast master transcription factor Runx2, we monitored Runx2
mRNA throughout this process. Surprisingly, Runx2 mRNA steadily declined
between days 1 and 14 (Fig. 11B). This demonstrates that Runx2 is negatively
regulated at the mRNA level throughout osteoblast development, in stark contrast to
the positive regulation of OC (Fig. 11A).

Runx2-mediated OC expression could be developmentally up-regulated through
translational control of Runx2 [Sudhakar et al., 2001], and so we measured Runx2
protein levels during MC3T3-E1 osteoblast differentiation. Remarkably, Runx2
protein levels also dramatically decreased as OC expression increased (Fig. 11C).
Thus, despite evidence for translational up-regulation of Runx2 immediately after
confluence (Fig. 11B and 1C, compare day 4 to day 1), the developmental up-
118
regulation of OC thereafter cannot be explained by Runx2 protein levels, and may be
accounted for by post-translational regulatory control of Runx2.

119
Fig. 45. Increased Runx2 occupancy, not expression or DNA-binding activity
parallels OC up-regulation during osteoblast maturation. MC3T3-E1 cells were
cultured in the presence of differentiation media for 14 days after confluence. Total
RNA, whole-cell protein extracts, and cross-linked chromatin were collected
concomitantly on days 1, 4, 8, 11, and 14. (A, B) OC and Runx2 mRNA were
measured by RT-qPCR and the results were expressed relative to Ribosomal protein
L10A mRNA levels. (C) Western blot analysis of Runx2. Coomassie-stained bands
are shown to demonstrate equal loading. (D) EMSA with an OSE2 probe was
performed on the indicated days in the absence (left) or presence (right) of Runx2
antibodies. Black arrowhead – Runx2 shift; white arrowhead – super-shift. (E)
Runx2 occupancy in vivo was measured by ChIP with Runx2-specific antibodies and
qPCR with primers directed against the OC (closed circles) or Insulin (open circles)
promoters. Values for each locus are expressed as a percentage of the signal
obtained from input DNA. (F) Histone H3 acetylation at the OC promoter was
measured by ChIP assay. Error bars are standard deviations of triplicate
measurements.
120
0
1
2
3
4
0
2
4
6
OC
Runx2
Relative mRNA Levels
0
0.2
0.4
0.6
0.8
OC
Insulin
Percent Input
A.
B.
D.
E.
C.
1 4 11 14 Day:
1 4 11 14 Day: 8
Runx2
Coomassie
Day
14
+ α-Runx2 Ab
11 8 4 14 11 8 4 1
Day 1
Day 11
Day 11
+Ab
-Ab
+Ab
G.
F.
AcH3 ChIP
0
0.5
1
1.5
2
2.5
13 57 9 11 13
OC (-285/-86)
Insulin (-160/+31)
% Input
121
Developmental up-regulation of osteocalcin is attributable to enhanced Runx2
occupancy.
Post-translational regulation of Runx2 may affect its activity by altering its ability to
interact with DNA [Qiao et al., 2004]. To address this possibility, we monitored
binding activity by EMSA with a Runx2-binding OSE2 probe [Luppen et al., 2003].
A specific complex that can be completely super-shifted by Runx2 antibodies first
appeared on Day 4 and diminished gradually until Day 14 (Fig. 11D) in a pattern
reminiscent of Runx2 total protein levels (Fig. 11C). This indicates that Runx2
DNA-binding activity, much like its mRNA and protein levels, decreases while OC
expression surges. Because the samples for measurement of RNA, protein, and
DNA binding activity were all collected concomitantly, it is unlikely that the
incongruity between OC expression and Runx2 is merely an experimental artifact.
Rather, OC up-regulation could occur via post-translational control of Runx2 that
does not necessarily affect DNA binding. Even though Runx2 protein is present and
appears to be fully capable of binding DNA as early as day 4, it may not be able to
access its genomic targets in vivo. To test this hypothesis, we monitored Runx2
interaction with the OSE2 element in its native chromatin environment via ChIP
assay of the OC promoter. Remarkably, we found that occupancy indeed increases
dramatically between Days 4 and 11 (Fig. 11D), the period during which Runx2
mRNA, protein and DNA-binding activity decrease in parallel cultures (Fig. 11B-D).
This suggests that Runx2 is blocked from binding to the OC promoter during early
stages of osteoblast differentiation. Likely, once Runx2 can access OC (and possibly
122
other genomic targets), its own transcription is down-regulated (Fig. 11B). Be that
as it may, our data support the notion that the developmental regulation of OC by
Runx2 is at the level of chromatin occupancy. Although the exact underlying
mechanism remains to be elucidated, it is unlikely that chromatin conformation is
responsible, since histone acetylation at the OC promoter does not seem to increase
as a function of time (Fig. 43F).
Fig. 46. Conventional ChIP confirmation of Runx2 binding sites discovered by ChIP-chip

Table V. Peaks Identified by Replicate ChIP-chip Experiments

A. Number of Peaks Identified by Each Experiment
Experiment FDR=0.001 FDR=0.005 FDR=0.01 FDR=0.05 FDR=0.1
1 (original) 64 209 351 1186 2070
2 (technical) 45 118 190 770 1522
3 (biological) 45 95 122 295 544
B. Number of Overlapping Peaks
Experiments FDR=0.001 FDR=0.005 FDR=0.01 FDR=0.05 FDR=0.1
1, 2 24 64 112 477 977
1, 3 3 8 9 37 86
2, 3 2 6 10 53 102
1, 2, 3 0 2 3 26 58

0
1
2
3
4
5
6
7
1
2
3
4
7
9
10
11
15
17
18
20
21
22
24
27
29
30
33
35
36
38
39
41
42
43
44
46
47
49
52
55
56
57
59
60
61
62
63
Oc
Ins
1
2
3
5
6
7
8
9
123
Developmental up-regulation of Runx2 occupancy at the novel target, Glt28d2.
To test whether the dynamic pattern of Runx2 genomic occupancy is specific to the
OC locus or shared by other targets, we initially searched for additional Runx2-
bound sites in the genome of MC3T3-E1 cells undergoing terminal differentiation.
To this end, late-stage cultures were subjected to Runx2 ChIP-chip analysis using a
NimbleGen array that tiles across 2.6% of the mouse genome. Despite a high false
discovery rate (see Table V and Fig. 44), we identified a novel locus occupied by
Runx2, which mapped 1.2-kb upstream from the transcription start site of
Glycosyltransferase 28 domain containing 1-like (Glt28d2), a gene with no known
function (Fig. 12A). Using qPCR with site-specific primers, we found that Runx2
occupancy at the Glt28d3 locus was as robust as the interaction with the OC locus
(Fig. 12B). We then performed an independent series of ChIP assays to measure
recruitment of Runx2 to the Glt28d2 locus as a function of time during the
development of the osteoblast phenotype in MC3T3-E1 cultures. Remarkably,
recruitment of Runx2 to the Glt28d2 locus dramatically increased after day 4. As in
Fig. 11E, occupancy of the OC locus was again developmentally controlled, and
almost perfectly paralleled that of Glt28d2 (Fig. 12C). These results suggest that the
developmental up-regulation of Runx2’s access to genomic targets may be a global
phenomenon, not just a local event occurring at the OC locus. Finally, we measured
Glt28d2 mRNA levels, and found that unlike OC, this gene’s expression declines as
a function of time (Fig. 45D). Thus, interaction of Runx2 with the Glt28d2 promoter
may result in the repression of Glt28d2 transcription, or it may be controlling the
124
expression of another gene’s expression via this site. Further work must be done to
test both of these possibilities.

Fig. 47. Identification of a novel Runx2-occupied genomic target and characterization of its occupancy
during osteoblast differentiation. (A) Runx2 ChIPs were prepared from two independent late-stage MC3T3-E1
osteoblast cultures and were each subjected to ChIP-chip analysis using NimbleGen chip #7 of the MM8 tiling
set. Results from one ChIP-chip experiment are shown for a reproducibly-enriched locus. Enrichment values,
based on hybridization intensity compared to input, are presented for twenty 50-bp probes (black squares) tiling a
2.1-kb fragment that encompasses the transcription start site (bent arrow) of the Glt28d2 gene (black rectangle).
Consensus Runx2 elements are shown as black triangles and the position of primers used to measure occupancy
by qPCR are shown as horizontal arrows. (B) Runx2 recruitment to the Glt28d2, as well as the OC (positive
control) and the insulin (negative control) loci, was assessed by qPCR analysis of the DNA used for ChIP-Chip.
(C) Presence of Runx2 at the Glt28d2, OC, and Insulin loci was measured by an independent series of ChIP
assays performed on the indicated days during development of the osteoblast phenotype. Values represent the
percentage of signal obtained from the respective input DNA samples. (D) Glt28d2 mRNA levels were measured
in differentiating MC3T3-E1 cells by RT-qPCR. Error bars are standard deviations of triplicate measurements.

5.4 Discussion
Expression of Runx2, a master transcription factor required for osteoblast
differentiation, is often considered an indicator of lineage progression. We observed

0
5
10
15
20
25
A.
B.
OC Glt28d2 Insulin
Enrichment
(Relative copy no.)
Enrichment
(Log 2 ChIP/Input)
0.0
3.6
Glt28d2
2,124 bp
1.8
C.
0
0.05
0.1
0.15
0.2
0.25
OC
Glt28d2
Insulin
Enrichment
(Percent Input)
1 4 11 14 Day: 8
0
0.5
1
1.5
2
Day: 1 4 8 11 14
Glt28d2
mRNA levels
D.
125
that Runx2 mRNA levels actually decrease in MC3T3-E1 pre-osteoblasts during the
two-week period after confluence, in which they secrete and subsequently mineralize
a bone-like ECM. Thus, while Runx2 mRNA is undoubtedly a good marker for
early commitment of mesenchymal stem cells to the osteoblast lineage [Ducy et al.,
1997], the present work suggests that it cannot be used to measure progression of the
osteoblast phenotype in committed cells. In contrast, our results confirm the
usefulness of OC mRNA as a marker for monitoring progression of the osteoblast
phenotype in committed mesenchymal cells. Measurement of both Runx2 and OC
mRNA is therefore an efficient approach to follow development of the osteoblast
phenotype, with the appearance of Runx2 transcripts marking early commitment, and
OC mRNA levels indicating the degree of differentiation in committed cells.

Our results reveal that mRNA is not the only Runx2 parameter that shares a dynamic
relationship with OC expression during osteoblast differentiation. For example, in
very early MC3T3-E1 cultures (day 1), when Runx2 mRNA levels are highest, the
absence of OC expression is attributable simply to lack of Runx2 protein (Western).
This disconnect between Runx2 mRNA and protein levels likely reflects a
translational block [Sudhakar et al., 2001]. The transcriptional silence of OC on day
4 cannot be attributed to such a block though, because Runx2 protein (Western) and
DNA-binding potential (EMSA) are in fact maximal at this point. Instead, it may be
explained by a block in recruitment, because Runx2 is essentially absent from the
OC promoter (ChIP). As the cultures continue to mature, Runx2 is recruited to this
126
locus; thus, increased occupancy is presumably responsible for the stimulation of OC
mRNA levels. Interestingly, during late stages of culture progression (day 14), OC
mRNA is maximal despite decreasing Runx2 occupancy at the OC locus. This
apparent discrepancy, which is not the focus of the present work, may reflect
stimulation of the OC promoter by transcription factors other than Runx2.

During early stages of culture progression, Runx2 was excluded from both the OC
and Glt28d2 promoters, despite peak protein levels and DNA-binding activity. This
may be due to several mechanisms: (i) Runx2 could be sequestered in the cytoplasm
[Kim et al., 2003]; (ii) Runx2 and its genomic targets could be partitioned to
different sub-nuclear compartments [Zaidi et al., 2001]; and (iii) The chromatin
structure at the OC and Glt28d2 loci could render them inaccessible to Runx2 [Shen
et al., 2003]. Future work is warranted to elucidate both the exclusion mechanism
and the signals that overcome it, thereby allowing recruitment of Runx2 to target
loci. Such signals may originate from the ECM, which appears in concert with the
increase in Runx2 occupancy (data not shown). They may affect Runx2 post-
translationally [Franceschi and Xiao, 2003], or they may induce the expression of
collaborating factors required for access of Runx2 to its genomic targets.

Increased Runx2 genomic occupancy is accompanied by decreased Runx2
expression. This may reflect negative auto-regulation [Drissi et al., 2000], or it may
be due to repression by other factors. Regardless, osteoblasts have adapted a system
127
for quenching the expression of Runx2 around the time at which it gains full access
to its genomic targets. Presumably, it is the attrition of Runx2 protein that eventually
leads to reduced occupancy. The implication of these findings is that there is only a
brief window of time during which the full transcriptional stimulatory ability of
Runx2 is unleashed. The need for such stringent regulatory control is consistent with
the severe consequences of Runx2 mis-regulation in vivo [Komori et al., 1997;
Maruyama et al., 2007; Otto et al., 1997].

In summary, we measured various aspects of Runx2 expression and function in
developing osteoblasts and discovered that Runx2 occupancy of genomic targets
(demonstrated by ChIP assays), rather than DNA-binding potential (EMSA) or
expression (RT-qPCR, Western analysis), parallels OC mRNA during progression of
the osteoblast phenotype in MC3T3-E1 cultures. Recruitment of Runx2 to its
genomic targets could be facilitated by alterations in chromatin structure, although
the near-identical developmental patterns of Runx2 occupancy at the OC and the
Glt28d2 loci favor the hypothesis that a collaborating factor(s) or post-translational
control of Runx2 itself serves the switch for allowing interaction with genomic
targets during osteoblast differentiation.
128
Chapter 8: Discussion

Genome-wide location analysis of Runx2 in osteoblasts
At the beginning of these studies, genome-wide location analysis was in its infancy.
While most researchers were still using ChIP to probe interactions between
transcription factors and their binding sites on a candidate basis, a handful of labs
were exploring its potential to uncover novel targets [Kurdistani and Grunstein,
2003; Ren et al., 2000; Weinmann and Farnham, 2002]. The efforts of these
pioneers, coupled with significant technological advances have helped genome-wide
location analysis to become an important tool for dissecting transcription factor
mechanisms today. Surprisingly, the bone community has been somewhat slow in
applying this valuable tool to the study of Runx2’s mechanism in osteoblasts.
Indeed, ours is the first lab that has endeavored to do so. Since it is unlikely that we
will also be the last, a discussion of the insights we have gained into genome-wide
location analysis of Runx2 in osteoblasts is warranted.

Successful genome-wide location analysis begins with a proper command of the
ChIP technique. Although ChIP is now fairly commonplace, few labs can claim to
have mastered it, especially in osteoblasts. Mastery of ChIP is proportional to the
level of enrichment one can achieve. Enrichment, in turn, is measured most
accurately by real-time PCR with locus-specific primers, comparing the
concentrations of known targets to those of non-targets. Although a number of loci
129
may be used for this purpose, the OSE2 element of the osteocalcin promoter and an
upstream element of the insulin promoter work well in our hands. Indeed, we
routinely achieve ~10-fold enrichment for OC vs. insulin in MC3T3-E1 cells,
although this depends largely on the timing of collection (Fig. 11E). The degree of
enrichment also depends greatly on several aspects of the technique itself, especially
fixation and sonication. The goals of these two important steps are to cross-link the
proteins and DNA that are in close proximity to one another, and to shear the cross-
linked chromatin down to ~500bp fragments, respectively. Failure to accomplish
either of these goals will result in an unsuccessful ChIP. Due to the heavy ECM
secreted by osteoblasts in general and MC3T3-E1 cells in particular, excessive
fixation can render the chromatin more resistant to shearing by sonication [Shen et
al., 2003]. Meanwhile, compensation by over-sonicating can result in destruction of
the epitope that is needed for immunoprecipitation. Thus, achieving optimal
enrichment involves finding the appropriate balance between under- and over-
crosslinking, as well as between under- and over-sonication. Even good enrichment
levels can be lost during an amplification step, which most ChIP-based location
analysis techniques require. This phenomenon, known as amplification bias, can
occur when the enriched locus is not amplified as efficiently as the negative control
locus, due to differing features such as GC content. Although it is not completely
avoidable, it can be limited by increasing the amount of starting material, so that
fewer rounds of amplification are required. Moreover, for reasons that are not
entirely clear, certain amplification methods tend to exaggerate the bias less than
130
others [O'Geen et al., 2006]. Regardless, it is critical to measure enrichment after
amplification to ascertain whether or not the amount of bias is acceptable.

Even when enrichment survives the amplification step, targets may be missed by
genome-wide location analysis. For example, neither CD nor ChIP-Chip could
detect the OSE2 element of the Osteocalcin promoter as a Runx2 binding site in
osteoblasts, despite the fact that this locus was enriched in the ChIP samples used for
these experiments, before and after amplification. For CD, this was expected due to
a technical aspect of the method, which limits its potential for target discovery.
Specifically, the ability to disclose a given target depends on the presence of two
AvaII recognition sites in the vicinity of that target. The genomic distribution of
recognition sequences for enzymes such as AvaII can be expected to result in many
sonication fragments containing two restriction sites; however, many sonication
fragments will not be long enough and targets will be missed. Indeed, this is likely
what happened for the OSE2 locus. Moreover, digestion by AvaII can be blocked by
CpG methylation, further limiting its potential for target discovery. Thus, ChIP
Display is only a ‘sampling method,’ and that it will not disclose all Runx2 genomic
targets. In contrast, ChIP-chip suffers from no such limitations, and has the potential
to uncover hundreds of targets [Deblois and Giguere, 2008]. The underlying cause
as to why we found so few is an enigma, although we suspect that the fault lies with
one of the downstream steps, such as sample labeling, hybridization, or array
scanning. In any case, as with CD, the paucity of targets likely reflects shortcomings
131
in the method’s sensitivity, rather than the actual biological situation. Given the
likelihood that many genomic targets of Runx2 in osteoblasts remain to be
discovered, further studies involving genome-wide location analysis are warranted.
In this regard, CD is a straightforward approach that can be used successfully to
identify Runx2 targets in osteoblasts. ChIP-chip, on the other hand, is much higher-
throughput and can potenially identify many more binding sites; however, it is not
without serious technical challenges.

Multiple levels of regulation result in poor correlations between Runx2 expression
and activity
In MC3T3-E1 cultures, we have observed significant variations in the trends of
Runx2 mRNA levels as a function of maturational status (compare Figs. 10a & 11b).
This may be due to primer differences (compare Tables I and II); however, it seems
unlikely, given that both primer sets targeted the major isoforms of Runx2 and had
roughly equal amplification efficiencies (data not shown). Rather than
compromising our results, this apparent contradiction serves to highlight a key point;
namely, major inferences about Runx2 function should not be drawn on the basis of
its mRNA levels alone. Not only did these vary greatly between experiments, they
generally did not correlate very well with OC mRNA, which was measured in
parallel (Figs. 7b, 8b, 9b, 10b, and 11a). OC is a classic Runx2 target gene whose
expression is directly influenced by Runx2 binding to a cis-acting element in its
proximal promoter region [Banerjee et al., 1997; Ducy and Karsenty, 1995; Ducy et
132
al., 1997; Merriman et al., 1995]. Thus, OC mRNA serves as a reliable indicator of
Runx2 activity. Indeed, it likely is more reliable than Runx2 mRNA, given the lack
of correlation between these two parameters. Interestingly, Runx2 protein and in
vitro DNA binding did not correlate well with OC mRNA either, suggesting that
these parameters may also poorly reflect Runx2 activity.

The general lack of correlation between Runx2 expression, in vitro DNA binding,
and target gene activity suggests that Runx2 is regulated at multiple levels via
independent mechanisms. This notion is also supported by a plethora of
experimental evidence. For example, at the transcriptional level, Runx2 is regulated
by a number of transcription factors that can stimulate or repress promoter activity.
This includes JunB [Lee et al., 2002], FosB [Zambotti et al., 2002], Smad3 [Alliston
et al., 2001], TCF-1 [Gaur et al., 2005], VD3 receptor [Drissi et al., 2002], ER [Tou
et al., 2001], USF2 [Wang et al., 2006], MSX2, DLX3, DLX5 [Hassan et al., 2006;
Lee et al., 2003; Lee et al., 2005], and Runx2 itself [Drissi et al., 2000].
Additionally, differential promoter usage [Geoffroy et al., 1998; Harada et al., 1999]
and alternative splicing [Terry et al., 2004; Xiao et al., 2001; Xiao et al., 1998] both
result in multiple Runx2 isoforms. These, in turn, may be subject to translation by
different mechanisms [Elango et al., 2006; Sudhakar et al., 2001; Xiao et al., 2003].

Post-translationally, Runx2 is subject to regulation via various mechanisms as well.
This includes control of its movement into the nucleus [Kim et al., 2003; Pockwinse
133
et al., 2006; Thirunavukkarasu et al., 1998], as well as its sub-nuclear localization
and association with the nuclear matrix [Harrington et al., 2002; Lindenmuth et al.,
1997; Zaidi et al., 2001; Zaidi et al., 2006]. Moreover, Runx2 undergoes several
covalent modifications, including phosphorylation [Afzal et al., 2005; Kim et al.,
2006; Phillips et al., 2006; Shui et al., 2003; Wang et al., 2002; Wee et al., 2002;
Xiao et al., 2000; Ziros et al., 2002], ubiquitination [Bellido et al., 2003; Kaneki et
al., 2006; Shen et al., 2006; Shen et al., 2006; Zhao et al., 2003], acetylation [Jeon et
al., 2006], and O-Linked β-N-acetylglucosamine (O-GlcNAc) modification [Kim et
al., 2007]. These, in turn, may affect Runx2 protein levels or its ability to bind
DNA. Runx2 DNA binding may also be affected by interactions with other proteins.
For example, it is enhanced by Cbfa β [Kundu et al., 2002; Yoshida et al., 2002] and
c-Fos/c-Jun [D'Alonzo et al., 2002], while it is suppressed by others, including
Peroxisome proliferator-activated receptor- [Jeon et al., 2003], Twist proteins 1 and
2 [Bialek et al., 2004], Nrf2 [Hinoi et al., 2006], and SOX9 [Zhou et al., 2006].

Regulation of Runx2 at multiple levels via independent mechanisms is consistent
with a stringent requirement for varying amounts of functional Runx2 in osteoblasts
throughout their development. This requirement, in turn, is reflected by the
devastating skeletal consequences of Runx2 mis-regulation (see Introduction).
Future work should address the mechanisms that govern the various levels of Runx2
regulatory control, including upstream signals and pathways. Understanding these
mechanisms more completely should help to clarify the etiology of various skeletal
134
diseases, as well as provide greater leverage for the use of Runx2 as a therapeutic
tool.

In summary, we have identified several novel genomic targets of Runx2 in
osteoblasts by location analysis. One of these mapped to the promoter of Tram2,
which may be involved in type I collagen synthesis. We provided evidence that
Runx2 regulates the expression of this gene in a cell-type specific and BMP-
dependent manner. Moreover, we showed that Runx2 stimulation of OC expression
during osteoblast development is attributable to increased promoter occupancy, and
that this occurs in spite of decreasing Runx2 expression and in vitro DNA binding
activity. Finally, we showed that Runx2 occupies at least one other newly-identified
genomic target in the same developmentally-regulated fashion.
135
References

Afzal F, Pratap J, Ito K, Ito Y, Stein JL, van Wijnen AJ, Stein GS, Lian JB, Javed A
(2005): Smad function and intranuclear targeting share a Runx2 motif required for
osteogenic lineage induction and BMP2 responsive transcription. J Cell Physiol
204:63-72.
Alliston T, Choy L, Ducy P, Karsenty G, Derynck R (2001): TGF-beta-induced
repression of CBFA1 by Smad3 decreases cbfa1 and osteocalcin expression and
inhibits osteoblast differentiation. Embo J 20:2254-72.
Ando T, Ichikawa J, Wako M, Hatsushika K, Watanabe Y, Sakuma M, Tasaka K,
Ogawa H, Hamada Y, Yagita H, Nakao A (2006): TWEAK/Fn14 interaction
regulates RANTES production, BMP-2-induced differentiation, and RANKL
expression in mouse osteoblastic MC3T3-E1 cells. Arthritis Res Ther 8:R146.
Aubin JE, Gupta AK, Zirngbl R, Rossant J (1996): Knockout mice lacking bone
sialoprotein expression have bone abnormalities. J. Bone Miner. Res. 11 (supplement
1):S102 (abstract).
Banerjee C, McCabe LR, Choi JY, Hiebert SW, Stein JL, Stein GS, Lian JB (1997):
Runt homology domain proteins in osteoblast differentiation: AML3/CBFA1 is a
major component of a bone-specific complex. J Cell Biochem 66:1-8.
Barski A, Frenkel B (2004): ChIP Display: novel method for identification of
genomic targets of transcription factors. Nucleic Acids Res 32:e104.
Barski A, Pregizer S, Frenkel B (2008): Identification of transcription factor target
genes by ChIP display. Methods Mol Biol 455:177-90.
Bellido T, Ali AA, Plotkin LI, Fu Q, Gubrij I, Roberson PK, Weinstein RS, O'Brien
CA, Manolagas SC, Jilka RL (2003): Proteasomal degradation of Runx2 shortens
parathyroid hormone-induced anti-apoptotic signaling in osteoblasts. A putative
explanation for why intermittent administration is needed for bone anabolism. J Biol
Chem 278:50259-72.
136
Bertaux K, Broux O, Chauveau C, Hardouin P, Jeanfils J, Devedjian JC (2006):
Runx2 regulates the expression of GNAS on SaOs-2 cells. Bone 38:943-50.
Bertaux K, Broux O, Chauveau C, Jeanfils J, Devedjian JC (2005): Identification of
CBFA1-regulated genes on SaOs-2 cells. J Bone Miner Metab 23:114-22.
Bialek P, Kern B, Yang X, Schrock M, Sosic D, Hong N, Wu H, Yu K, Ornitz DM,
Olson EN, Justice MJ, Karsenty G (2004): A twist code determines the onset of
osteoblast differentiation. Dev Cell 6:423-35.
Bieda M, Xu X, Singer MA, Green R, Farnham PJ (2006): Unbiased location
analysis of E2F1-binding sites suggests a widespread role for E2F1 in the human
genome. Genome Res 16:595-605.
Bryan J, Munoz A, Zhang X, Dufer M, Drews G, Krippeit-Drews P, Aguilar-Bryan
L (2007): ABCC8 and ABCC9: ABC transporters that regulate K(+) channels.
Pflugers Arch 453:703-18.
Bucay N, Sarosi I, Dunstan CR, Morony S, Tarpley J, Capparelli C, Scully S, Tan
HL, Xu W, Lacey DL, Boyle WJ, Simonet WS (1998): osteoprotegerin-deficient
mice develop early onset osteoporosis and arterial calcification. Genes Dev 12:1260-
8.
Byers BA, Garcia AJ (2004): Exogenous Runx2 expression enhances in vitro
osteoblastic differentiation and mineralization in primary bone marrow stromal cells.
Tissue Eng 10:1623-32.
Byers BA, Guldberg RE, Garcia AJ (2004): Synergy between genetic and tissue
engineering: Runx2 overexpression and in vitro construct development enhance in
vivo mineralization. Tissue Eng 10:1757-66.
Byers BA, Pavlath GK, Murphy TJ, Karsenty G, Garcia AJ (2002): Cell-type-
dependent up-regulation of in vitro mineralization after overexpression of the
osteoblast-specific transcription factor Runx2/Cbfal. J Bone Miner Res 17:1931-44.

137
Cawley S, Bekiranov S, Ng HH, Kapranov P, Sekinger EA, Kampa D, Piccolboni A,
Sementchenko V, Cheng J, Williams AJ, Wheeler R, Wong B, Drenkow J,
Yamanaka M, Patel S, Brubaker S, Tammana H, Helt G, Struhl K, Gingeras TR
(2004): Unbiased mapping of transcription factor binding sites along human
chromosomes 21 and 22 points to widespread regulation of noncoding RNAs. Cell
116:499-509.
Chen D, Zhao M, Mundy GR (2004): Bone morphogenetic proteins. Growth Factors
22:233-41.
Chen J, Sadowski I (2005): Identification of the mismatch repair genes PMS2 and
MLH1 as p53 target genes by using serial analysis of binding elements. Proc Natl
Acad Sci U S A 102:4813-8.
Choi KY, Kim HJ, Lee MH, Kwon TG, Nah HD, Furuichi T, Komori T, Nam SH,
Kim YJ, Kim HJ, Ryoo HM (2005): Runx2 regulates FGF2-induced Bmp2
expression during cranial bone development. Dev Dyn 233:115-21.
Colnot C, Fowlis D, Ripoche MA, Bouchaert I, Poirier F (1998): Embryonic
implantation in galectin 1/galectin 3 double mutant mice. Dev Dyn 211:306-13.
Cui CB, Cooper LF, Yang X, Karsenty G, Aukhil I (2003): Transcriptional
coactivation of bone-specific transcription factor Cbfa1 by TAZ. Mol Cell Biol
23:1004-13.
D'Alonzo RC, Selvamurugan N, Karsenty G, Partridge NC (2002): Physical
interaction of the activator protein-1 factors c-Fos and c-Jun with Cbfa1 for
collagenase-3 promoter activation. J Biol Chem 277:816-22.
Deblois G, Giguere V (2008): Nuclear Receptor Location Analyses in Mammalian
Genomes: From Gene Regulation to Regulatory Networks. Mol Endocrinol.
Deregowski V, Gazzerro E, Priest L, Rydziel S, Canalis E (2006): Notch 1
overexpression inhibits osteoblastogenesis by suppressing Wnt/beta-catenin but not
bone morphogenetic protein signaling. J Biol Chem 281:6203-10.
138
Desbois C, Hogue DA, Karsenty G (1994): The mouse osteocalcin gene cluster
contains three genes with two separate spatial and temporal patterns of expression. J
Biol Chem 269:1183-90.
Doecke JD, Day CJ, Stephens AS, Carter SL, van Daal A, Kotowicz MA, Nicholson
GC, Morrison NA (2006): Association of functionally different RUNX2 P2 promoter
alleles with BMD. J Bone Miner Res 21:265-73.
Drissi H, Luc Q, Shakoori R, Chuva De Sousa Lopes S, Choi JY, Terry A, Hu M,
Jones S, Neil JC, Lian JB, Stein JL, Van Wijnen AJ, Stein GS (2000):
Transcriptional autoregulation of the bone related CBFA1/RUNX2 gene. J Cell
Physiol 184:341-50.
Drissi H, Pouliot A, Koolloos C, Stein JL, Lian JB, Stein GS, van Wijnen AJ (2002):
1,25-(OH)2-vitamin D3 suppresses the bone-related Runx2/Cbfa1 gene promoter.
Exp Cell Res 274:323-33.
Ducy P, Desbois C, Boyce B, Pinero G, Story B, Dunstan C, Smith E, Bonadio J,
Goldstein S, Gundberg C, Bradley A, Karsenty G (1996): Increased bone formation
in osteocalcin-deficient mice. Nature 382:448-52.
Ducy P, Karsenty G (1995): Two distinct osteoblast-specific cis-acting elements
control expression of a mouse osteocalcin gene. Mol Cell Biol 15:1858-69.
Ducy P, Starbuck M, Priemel M, Shen J, Pinero G, Geoffroy V, Amling M, Karsenty
G (1999): A Cbfa1-dependent genetic pathway controls bone formation beyond
embryonic development. Genes Dev 13:1025-36.
Ducy P, Zhang R, Geoffroy V, Ridall AL, Karsenty G (1997): Osf2/Cbfa1: a
transcriptional activator of osteoblast differentiation. Cell 89:747-54.
Elango N, Li Y, Shivshankar P, Katz MS (2006): Expression of RUNX2 isoforms:
involvement of cap-dependent and cap-independent mechanisms of translation. J
Cell Biochem 99:1108-21.

139
Enomoto H, Shiojiri S, Hoshi K, Furuichi T, Fukuyama R, Yoshida CA, Kanatani N,
Nakamura R, Mizuno A, Zanma A, Yano K, Yasuda H, Higashio K, Takada K,
Komori T (2003): Induction of osteoclast differentiation by Runx2 through receptor
activator of nuclear factor-kappa B ligand (RANKL) and osteoprotegerin regulation
and partial rescue of osteoclastogenesis in Runx2-/- mice by RANKL transgene. J
Biol Chem 278:23971-7.
Ermakov S, Malkin I, Kobyliansky E, Livshits G (2006): Variation in femoral length
is associated with polymorphisms in RUNX2 gene. Bone 38:199-205.
Fen JQ, Zhang J, Dallas SL, Lu Y, Chen S, Tan X, Owen M, Harris SE, MacDougall
M (2002): Dentin matrix protein 1, a target molecule for Cbfa1 in bone, is a unique
bone marker gene. J Bone Miner Res 17:1822-31.
Franceschi RT, Xiao G (2003): Regulation of the osteoblast-specific transcription
factor, Runx2: responsiveness to multiple signal transduction pathways. J Cell
Biochem 88:446-54.
Frenkel B, Capparelli C, Van Auken M, Baran D, Bryan J, Stein JL, Stein GS, Lian
JB (1997): Activity of the osteocalcin promoter in skeletal sites of transgenic mice
and during osteoblast differentiation in bone marrow-derived stromal cell cultures:
effects of age and sex. Endocrinology 138:2109-16.
Frenkel B, Montecino M, Green J, Aslam F, Desai R, Banerjee C, Stein JL, Lian JB,
Stein GS (1996): Basal and vitamin D-responsive activity of the rat osteocalcin
promoter in stably transfected osteosarcoma cells: requirement of upstream
sequences for control by the proximal regulatory domain. Endocrinology 137:1080-
88.
Galindo M, Pratap J, Young DW, Hovhannisyan H, Im HJ, Choi JY, Lian JB, Stein
JL, Stein GS, van Wijnen AJ (2005): The bone-specific expression of Runx2
oscillates during the cell cycle to support a G1-related antiproliferative function in
osteoblasts. J Biol Chem 280:20274-85.
Gao YH, Shinki T, Yuasa T, Kataoka-Enomoto H, Komori T, Suda T, Yamaguchi A
(1998): Potential role of cbfa1, an essential transcriptional factor for osteoblast
differentiation, in osteoclastogenesis: regulation of mRNA expression of osteoclast
differentiation factor (ODF). Biochem Biophys Res Commun 252:697-702.
140
Gaur T, Lengner CJ, Hovhannisyan H, Bhat RA, Bodine PV, Komm BS, Javed A,
van Wijnen AJ, Stein JL, Stein GS, Lian JB (2005): Canonical WNT signaling
promotes osteogenesis by directly stimulating Runx2 gene expression. J Biol Chem
280:33132-40.
Geoffroy V, Corral DA, Zhou L, Lee B, Karsenty G (1998): Genomic organization,
expression of the human CBFA1 gene, and evidence for an alternative splicing event
affecting protein function. Mamm Genome 9:54-7.
Geoffroy V, Ducy P, Karsenty G (1995): A PEBP2 alpha/AML-1-related factor
increases osteocalcin promoter activity through its binding to an osteoblast-specific
cis-acting element. J Biol Chem 270:30973-9.
Geoffroy V, Kneissel M, Fournier B, Boyde A, Matthias P (2002): High bone
resorption in adult aging transgenic mice overexpressing cbfa1/runx2 in cells of the
osteoblastic lineage. Mol Cell Biol 22:6222-33.
Gersbach CA, Byers BA, Pavlath GK, Garcia AJ (2004): Runx2/Cbfa1 stimulates
transdifferentiation of primary skeletal myoblasts into a mineralizing osteoblastic
phenotype. Exp Cell Res 300:406-17.
Gersbach CA, Byers BA, Pavlath GK, Guldberg RE, Garcia AJ (2004):
Runx2/Cbfa1-genetically engineered skeletal myoblasts mineralize collagen
scaffolds in vitro. Biotechnol Bioeng 88:369-78.
Gersbach CA, Le Doux JM, Guldberg RE, Garcia AJ (2006): Inducible regulation of
Runx2-stimulated osteogenesis. Gene Ther 13:873-82.
Guicheux J, Lemonnier J, Ghayor C, Suzuki A, Palmer G, Caverzasio J (2003):
Activation of p38 mitogen-activated protein kinase and c-Jun-NH2-terminal kinase
by BMP-2 and their implication in the stimulation of osteoblastic cell differentiation.
J Bone Miner Res 18:2060-8.
Guo M, Jan LY, Jan YN (1996): Control of daughter cell fates during asymmetric
division: interaction of Numb and Notch. Neuron 17:27-41.
141
Gutierrez S, Javed A, Tennant DK, van Rees M, Montecino M, Stein GS, Stein JL,
Lian JB (2002): CCAAT/enhancer-binding proteins (C/EBP) beta and delta activate
osteocalcin gene transcription and synergize with Runx2 at the C/EBP element to
regulate bone-specific expression. J Biol Chem 277:1316-23.
Hanai J, Chen LF, Kanno T, Ohtani-Fujita N, Kim WY, Guo WH, Imamura T,
Ishidou Y, Fukuchi M, Shi MJ, Stavnezer J, Kawabata M, Miyazono K, Ito Y
(1999): Interaction and functional cooperation of PEBP2/CBF with Smads.
Synergistic induction of the immunoglobulin germline Calpha promoter. J Biol
Chem 274:31577-82.
Harada H, Tagashira S, Fujiwara M, Ogawa S, Katsumata T, Yamaguchi A, Komori
T, Nakatsuka M (1999): Cbfa1 isoforms exert functional differences in osteoblast
differentiation. J Biol Chem 274:6972-8.
Harrington KS, Javed A, Drissi H, McNeil S, Lian JB, Stein JL, Van Wijnen AJ,
Wang YL, Stein GS (2002): Transcription factors RUNX1/AML1 and
RUNX2/Cbfa1 dynamically associate with stationary subnuclear domains. J Cell Sci
115:4167-76.
Hassan MQ, Javed A, Morasso MI, Karlin J, Montecino M, van Wijnen AJ, Stein
GS, Stein JL, Lian JB (2004): Dlx3 transcriptional regulation of osteoblast
differentiation: temporal recruitment of Msx2, Dlx3, and Dlx5 homeodomain
proteins to chromatin of the osteocalcin gene. Mol Cell Biol 24:9248-61.
Hassan MQ, Tare RS, Lee SH, Mandeville M, Morasso MI, Javed A, van Wijnen AJ,
Stein JL, Stein GS, Lian JB (2006): BMP2 commitment to the osteogenic lineage
involves activation of Runx2 by DLX3 and a homeodomain transcriptional network.
J Biol Chem 281:40515-26.
Hecht J, Seitz V, Urban M, Wagner F, Robinson PN, Stiege A, Dieterich C, Kornak
U, Wilkening U, Brieske N, Zwingman C, Kidess A, Stricker S, Mundlos S (2007):
Detection of novel skeletogenesis target genes by comprehensive analysis of a
Runx2(-/-) mouse model. Gene Expr Patterns 7:102-12.
Hinoi E, Fujimori S, Wang L, Hojo H, Uno K, Yoneda Y (2006): Nrf2 negatively
regulates osteoblast differentiation via interfering with Runx2-dependent
transcriptional activation. J Biol Chem 281:18015-24.
142
Horak CE, Snyder M (2002): ChIP-chip: a genomic approach for identifying
transcription factor binding sites. Methods Enzymol 350:469-83.
Impey S, McCorkle SR, Cha-Molstad H, Dwyer JM, Yochum GS, Boss JM,
McWeeney S, Dunn JJ, Mandel G, Goodman RH (2004): Defining the CREB
regulon: a genome-wide analysis of transcription factor regulatory regions. Cell
119:1041-54.
Inada M, Wang Y, Byrne MH, Rahman MU, Miyaura C, Lopez-Otin C, Krane SM
(2004): Critical roles for collagenase-3 (Mmp13) in development of growth plate
cartilage and in endochondral ossification. Proc Natl Acad Sci U S A 101:17192-7.
Irizarry RA, Warren D, Spencer F, Kim IF, Biswal S, Frank BC, Gabrielson E,
Garcia JG, Geoghegan J, Germino G, Griffin C, Hilmer SC, Hoffman E, Jedlicka
AE, Kawasaki E, Martinez-Murillo F, Morsberger L, Lee H, Petersen D,
Quackenbush J, Scott A, Wilson M, Yang Y, Ye SQ, Yu W (2005): Multiple-
laboratory comparison of microarray platforms. Nat Methods 2:345-50.
Jariwala U, Prescott J, Jia L, Barski A, Pregizer S, Cogan JP, Arasheben A, Tilley
WD, Scher HI, Gerald WL, Buchanan G, Coetzee GA, Frenkel B (2007):
Identification of novel androgen receptor target genes in prostate cancer. Mol Cancer
6:39.
Javed A, Barnes GL, Jasanya BO, Stein JL, Gerstenfeld L, Lian JB, Stein GS (2001):
runt homology domain transcription factors (Runx, Cbfa, and AML) mediate
repression of the bone sialoprotein promoter: evidence for promoter context-
dependent activity of Cbfa proteins. Mol Cell Biol 21:2891-905.
Javed A, Guo B, Hiebert S, Choi JY, Green J, Zhao SC, Osborne MA, Stifani S,
Stein JL, Lian JB, van Wijnen AJ, Stein GS (2000): Groucho/TLE/R-esp proteins
associate with the nuclear matrix and repress RUNX
(CBF(alpha)/AML/PEBP2(alpha)) dependent activation of tissue-specific gene
transcription. J Cell Sci 113 (Pt 12):2221-31.
Jeon EJ, Lee KY, Choi NS, Lee MH, Kim HN, Jin YH, Ryoo HM, Choi JY, Yoshida
M, Nishino N, Oh BC, Lee KS, Lee YH, Bae SC (2006): Bone morphogenetic
protein-2 stimulates Runx2 acetylation. J Biol Chem 281:16502-11.
143
Jeon MJ, Kim JA, Kwon SH, Kim SW, Park KS, Park SW, Kim SY, Shin CS
(2003): Activation of peroxisome proliferator-activated receptor-gamma inhibits the
Runx2-mediated transcription of osteocalcin in osteoblasts. J Biol Chem 278:23270-
7.
Ji C, Casinghino S, Chang DJ, Chen Y, Javed A, Ito Y, Hiebert SW, Lian JB, Stein
GS, McCarthy TL, Centrella M (1998): CBFa(AML/PEBP2)-related elements in the
TGF-beta type I receptor promoter and expression with osteoblast differentiation. J
Cell Biochem 69:353-63.
Jia L, Berman BP, Jariwala U, Yan X, Cogan JP, Walters A, Chen T, Buchanan G,
Frenkel B, Coetzee GA (2008): Genomic Androgen Receptor-Occupied Regions
with Different Functions, Defined by Histone Acetylation, Coregulators and
Transcriptional Capacity. Submitted Manuscript.
Jia L, Kim J, Shen H, Clark PE, Tilley WD, Coetzee GA (2003): Androgen receptor
activity at the prostate specific antigen locus: steroidal and non-steroidal
mechanisms. Mol Cancer Res 1:385-92.
Jimenez MJ, Balbin M, Lopez JM, Alvarez J, Komori T, Lopez-Otin C (1999):
Collagenase 3 is a target of Cbfa1, a transcription factor of the runt gene family
involved in bone formation. Mol Cell Biol 19:4431-42.
Juttner KV, Perry MJ (2007): High-dose estrogen-induced osteogenesis is decreased
in aged RUNX2(+/-) mice. Bone 41:25-32.
Kahler RA, Westendorf JJ (2003): Lymphoid enhancer factor-1 and beta-catenin
inhibit Runx2-dependent transcriptional activation of the osteocalcin promoter. J
Biol Chem 278:11937-44.
Kanatani N, Fujita T, Fukuyama R, Liu W, Yoshida CA, Moriishi T, Yamana K,
Miyazaki T, Toyosawa S, Komori T (2006): Cbf beta regulates Runx2 function
isoform-dependently in postnatal bone development. Dev Biol 296:48-61.
Kaneki H, Guo R, Chen D, Yao Z, Schwarz EM, Zhang YE, Boyce BF, Xing L
(2006): Tumor necrosis factor promotes Runx2 degradation through up-regulation of
Smurf1 and Smurf2 in osteoblasts. J Biol Chem 281:4326-33.
144
Kang JS, Alliston T, Delston R, Derynck R (2005): Repression of Runx2 function by
TGF-beta through recruitment of class II histone deacetylases by Smad3. Embo J
24:2543-55.
Katagiri T, Yamaguchi A, Komaki M, Abe E, Takahashi N, Ikeda T, Rosen V,
Wozney JM, Fujisawa-Sehara A, Suda T (1994): Bone morphogenetic protein-2
converts the differentiation pathway of C2C12 myoblasts into the osteoblast lineage.
J Cell Biol 127:1755-66.
Kawahata H, Kikkawa T, Higashibata Y, Sakuma T, Huening M, Sato M, Sugimoto
M, Kuriyama K, Terai K, Kitamura Y, Nomura S (2003): Enhanced expression of
Runx2/PEBP2alphaA/CBFA1/AML3 during fracture healing. J Orthop Sci 8:102-8.
Kern B, Shen J, Starbuck M, Karsenty G (2001): Cbfa1 contributes to the osteoblast-
specific expression of type I collagen genes. J Biol Chem 276:7101-7.
Kim BG, Kim HJ, Park HJ, Kim YJ, Yoon WJ, Lee SJ, Ryoo HM, Cho JY (2006):
Runx2 phosphorylation induced by fibroblast growth factor-2/protein kinase C
pathways. Proteomics 6:1166-74.
Kim J, Bhinge AA, Morgan XC, Iyer VR (2005): Mapping DNA-protein interactions
in large genomes by sequence tag analysis of genomic enrichment. Nat Methods
2:47-53.
Kim S, Koga T, Isobe M, Kern BE, Yokochi T, Chin YE, Karsenty G, Taniguchi T,
Takayanagi H (2003): Stat1 functions as a cytoplasmic attenuator of Runx2 in the
transcriptional program of osteoblast differentiation. Genes Dev 17:1979-91.
Kim SH, Kim YH, Song M, An SH, Byun HY, Heo K, Lim S, Oh YS, Ryu SH, Suh
PG (2007): O-GlcNAc modification modulates the expression of osteocalcin via
OSE2 and Runx2. Biochem Biophys Res Commun 362:325-9.
Kobayashi H, Gao Y, Ueta C, Yamaguchi A, Komori T (2000): Multilineage
differentiation of Cbfa1-deficient calvarial cells in vitro. Biochem Biophys Res
Commun 273:630-6.
145
Kojima H, Uemura T (2005): Strong and rapid induction of osteoblast differentiation
by Cbfa1/Til-1 overexpression for bone regeneration. J Biol Chem 280:2944-53.
Komori T, Yagi H, Nomura S, Yamaguchi A, Sasaki K, Deguchi K, Shimizu Y,
Bronson RT, Gao YH, Inada M, Sato M, Okamoto R, Kitamura Y, Yoshiki S,
Kishimoto T (1997): Targeted disruption of Cbfa1 results in a complete lack of bone
formation owing to maturational arrest of osteoblasts. Cell 89:755-64.
Kundu M, Javed A, Jeon JP, Horner A, Shum L, Eckhaus M, Muenke M, Lian JB,
Yang Y, Nuckolls GH, Stein GS, Liu PP (2002): Cbfbeta interacts with Runx2 and
has a critical role in bone development. Nat Genet 32:639-44.
Kurdistani SK, Grunstein M (2003): In vivo protein-protein and protein-DNA
crosslinking for genomewide binding microarray. Methods 31:90-5.
Lambertini E, Penolazzi L, Tavanti E, Schincaglia GP, Zennaro M, Gambari R, Piva
R (2007): Human estrogen receptor alpha gene is a target of Runx2 transcription
factor in osteoblasts. Exp Cell Res 313:1548-60.
Lee JS, Thomas DM, Gutierrez G, Carty SA, Yanagawa S, Hinds PW (2006): HES1
cooperates with pRb to activate RUNX2-dependent transcription. J Bone Miner Res
21:921-33.
Lee KS, Hong SH, Bae SC (2002): Both the Smad and p38 MAPK pathways play a
crucial role in Runx2 expression following induction by transforming growth factor-
beta and bone morphogenetic protein. Oncogene 21:7156-63.
Lee KS, Kim HJ, Li QL, Chi XZ, Ueta C, Komori T, Wozney JM, Kim EG, Choi
JY, Ryoo HM, Bae SC (2000): Runx2 is a common target of transforming growth
factor beta1 and bone morphogenetic protein 2, and cooperation between Runx2 and
Smad5 induces osteoblast-specific gene expression in the pluripotent mesenchymal
precursor cell line C2C12. Mol Cell Biol 20:8783-92.

146
Lee MH, Javed A, Kim HJ, Shin HI, Gutierrez S, Choi JY, Rosen V, Stein JL, van
Wijnen AJ, Stein GS, Lian JB, Ryoo HM (1999): Transient upregulation of CBFA1
in response to bone morphogenetic protein-2 and transforming growth factor beta1 in
C2C12 myogenic cells coincides with suppression of the myogenic phenotype but is
not sufficient for osteoblast differentiation. J Cell Biochem 73:114-25.
Lee MH, Kim YJ, Kim HJ, Park HD, Kang AR, Kyung HM, Sung JH, Wozney JM,
Kim HJ, Ryoo HM (2003): BMP-2-induced Runx2 expression is mediated by Dlx5,
and TGF-beta 1 opposes the BMP-2-induced osteoblast differentiation by
suppression of Dlx5 expression. J Biol Chem 278:34387-94.
Lee MH, Kim YJ, Yoon WJ, Kim JI, Kim BG, Hwang YS, Wozney JM, Chi XZ,
Bae SC, Choi KY, Cho JY, Choi JY, Ryoo HM (2005): Dlx5 specifically regulates
Runx2 type II expression by binding to homeodomain-response elements in the
Runx2 distal promoter. J Biol Chem 280:35579-87.
Lee NK, Sowa H, Hinoi E, Ferron M, Ahn JD, Confavreux C, Dacquin R, Mee PJ,
McKee MD, Jung DY, Zhang Z, Kim JK, Mauvais-Jarvis F, Ducy P, Karsenty G
(2007): Endocrine regulation of energy metabolism by the skeleton. Cell 130:456-69.
Lian JB, Gundberg CM (1988): Osteocalcin. Biochemical considerations and clinical
applications. Clin Orthop Relat Res:267-91.
Lindenmuth DM, van Wijnen AJ, Hiebert S, Stein JL, Lian JB, Stein GS (1997):
Subcellular partitioning of transcription factors during osteoblast differentiation:
developmental association of the AML/CBF alpha/PEBP2 alpha-related transcription
factor-NMP-2 with the nuclear matrix. J Cell Biochem 66:123-32.
Liu CJ, Chang E, Yu J, Carlson CS, Prazak L, Yu XP, Ding B, Lengyel P, Di Cesare
PE (2005): The interferon-inducible p204 protein acts as a transcriptional coactivator
of Cbfa1 and enhances osteoblast differentiation. J Biol Chem 280:2788-96.
Liu T, Gao Y, Sakamoto K, Minamizato T, Furukawa K, Tsukazaki T, Shibata Y,
Bessho K, Komori T, Yamaguchi A (2007): BMP-2 promotes differentiation of
osteoblasts and chondroblasts in Runx2-deficient cell lines. J Cell Physiol 211:728-
35.
147
Liu W, Toyosawa S, Furuichi T, Kanatani N, Yoshida C, Liu Y, Himeno M, Narai S,
Yamaguchi A, Komori T (2001): Overexpression of Cbfa1 in osteoblasts inhibits
osteoblast maturation and causes osteopenia with multiple fractures. J Cell Biol
155:157-66.
Loh YH, Wu Q, Chew JL, Vega VB, Zhang W, Chen X, Bourque G, George J,
Leong B, Liu J, Wong KY, Sung KW, Lee CW, Zhao XD, Chiu KP, Lipovich L,
Kuznetsov VA, Robson P, Stanton LW, Wei CL, Ruan Y, Lim B, Ng HH (2006):
The Oct4 and Nanog transcription network regulates pluripotency in mouse
embryonic stem cells. Nat Genet 38:431-40.
Luan Y, Yu XP, Xu K, Ding B, Yu J, Huang Y, Yang N, Lengyel P, Di Cesare PE,
Liu CJ (2007): The retinoblastoma protein is an essential mediator of osteogenesis
that links the p204 protein to the Cbfa1 transcription factor thereby increasing its
activity. J Biol Chem 282:16860-70.
Luppen CA, Leclerc N, Noh T, Barski A, Khokhar A, Boskey AL, Smith E, Frenkel
B (2003): Brief bone morphogenetic protein 2 treatment of glucocorticoid-inhibited
MC3T3-E1 osteoblasts rescues commitment-associated cell cycle and mineralization
without alteration of Runx2. J Biol Chem 278:44995-5003.
Maruyama Z, Yoshida CA, Furuichi T, Amizuka N, Ito M, Fukuyama R, Miyazaki
T, Kitaura H, Nakamura K, Fujita T, Kanatani N, Moriishi T, Yamana K, Liu W,
Kawaguchi H, Nakamura K, Komori T (2007): Runx2 determines bone maturity and
turnover rate in postnatal bone development and is involved in bone loss in estrogen
deficiency. Dev Dyn 236:1876-90.
Matheny CJ, Speck ME, Cushing PR, Zhou Y, Corpora T, Regan M, Newman M,
Roudaia L, Speck CL, Gu TL, Griffey SM, Bushweller JH, Speck NA (2007):
Disease mutations in RUNX1 and RUNX2 create nonfunctional, dominant-negative,
or hypomorphic alleles. Embo J 26:1163-75.
McCarthy TL, Chang WZ, Liu Y, Centrella M (2003): Runx2 integrates estrogen
activity in osteoblasts. J Biol Chem 278:43121-9.
McCarthy TL, Ji C, Chen Y, Kim KK, Imagawa M, Ito Y, Centrella M (2000): Runt
domain factor (Runx)-dependent effects on CCAAT/ enhancer-binding protein delta
expression and activity in osteoblasts. J Biol Chem 275:21746-53.
148
Merriman HL, van Wijnen AJ, Hiebert S, Bidwell JP, Fey E, Lian J, Stein J, Stein
GS (1995): The tissue-specific nuclear matrix protein, NMP-2, is a member of the
AML/CBF/PEBP2/runt domain transcription factor family: interactions with the
osteocalcin gene promoter. Biochemistry 34:13125-32.
Meyers S, Downing JR, Hiebert SW (1993): Identification of AML-1 and the (8;21)
translocation protein (AML-1/ETO) as sequence-specific DNA-binding proteins: the
runt homology domain is required for DNA binding and protein-protein interactions.
Mol Cell Biol 13:6336-45.
Mizuno A, Amizuka N, Irie K, Murakami A, Fujise N, Kanno T, Sato Y, Nakagawa
N, Yasuda H, Mochizuki S, Gomibuchi T, Yano K, Shima N, Washida N, Tsuda E,
Morinaga T, Higashio K, Ozawa H (1998): Severe osteoporosis in mice lacking
osteoclastogenesis inhibitory factor/osteoprotegerin. Biochem Biophys Res Commun
247:610-5.
Nishimura R, Hata K, Harris SE, Ikeda F, Yoneda T (2002): Core-binding factor
alpha 1 (Cbfa1) induces osteoblastic differentiation of C2C12 cells without
interactions with Smad1 and Smad5. Bone 31:303-12.
Nishio Y, Dong Y, Paris M, O'Keefe RJ, Schwarz EM, Drissi H (2006): Runx2-
mediated regulation of the zinc finger Osterix/Sp7 gene. Gene 372:62-70.
O'Geen H, Nicolet CM, Blahnik K, Green R, Farnham PJ (2006): Comparison of
sample preparation methods for ChIP-chip assays. Biotechniques 41:577-80.
Ogawa E, Maruyama M, Kagoshima H, Inuzuka M, Lu J, Satake M, Shigesada K,
Ito Y (1993): PEBP2/PEA2 represents a family of transcription factors homologous
to the products of the Drosophila runt gene and the human AML1 gene. Proc Natl
Acad Sci U S A 90:6859-63.
Otto F, Kanegane H, Mundlos S (2002): Mutations in the RUNX2 gene in patients
with cleidocranial dysplasia. Hum Mutat 19:209-16.

149
Otto F, Thornell AP, Crompton T, Denzel A, Gilmour KC, Rosewell IR, Stamp GW,
Beddington RS, Mundlos S, Olsen BR, Selby PB, Owen MJ (1997): Cbfa1, a
candidate gene for cleidocranial dysplasia syndrome, is essential for osteoblast
differentiation and bone development. Cell 89:765-71.
Owen TA, Aronow M, Shalhoub V, Barone LM, Wilming L, Tassinari MS, Kennedy
MB, Pockwinse S, Lian JB, Stein GS (1990): Progressive development of the rat
osteoblast phenotype in vitro: reciprocal relationships in expression of genes
associated with osteoblast proliferation and differentiation during formation of the
bone extracellular matrix. J Cell Physiol 143:420-30.
Paredes R, Arriagada G, Cruzat F, Villagra A, Olate J, Zaidi K, van Wijnen A, Lian
JB, Stein GS, Stein JL, Montecino M (2004): Bone-specific transcription factor
Runx2 interacts with the 1alpha,25-dihydroxyvitamin D3 receptor to up-regulate rat
osteocalcin gene expression in osteoblastic cells. Mol Cell Biol 24:8847-61.
Pelletier N, Champagne N, Stifani S, Yang XJ (2002): MOZ and MORF histone
acetyltransferases interact with the Runt-domain transcription factor Runx2.
Oncogene 21:2729-40.
Petersen PH, Tang H, Zou K, Zhong W (2006): The enigma of the numb-Notch
relationship during mammalian embryogenesis. Dev Neurosci 28:156-68.
Phillips JE, Gersbach CA, Wojtowicz AM, Garcia AJ (2006): Glucocorticoid-
induced osteogenesis is negatively regulated by Runx2/Cbfa1 serine
phosphorylation. J Cell Sci 119:581-91.
Phillips JE, Guldberg RE, Garcia AJ (2007): Dermal fibroblasts genetically modified
to express Runx2/Cbfa1 as a mineralizing cell source for bone tissue engineering.
Tissue Eng 13:2029-40.
Pockwinse SM, Rajgopal A, Young DW, Mujeeb KA, Nickerson J, Javed A, Redick
S, Lian JB, van Wijnen AJ, Stein JL, Stein GS, Doxsey SJ (2006): Microtubule-
dependent nuclear-cytoplasmic shuttling of Runx2. J Cell Physiol 206:354-62.
150
Pratap J, Galindo M, Zaidi SK, Vradii D, Bhat BM, Robinson JA, Choi JY, Komori
T, Stein JL, Lian JB, Stein GS, van Wijnen AJ (2003): Cell growth regulatory role of
Runx2 during proliferative expansion of preosteoblasts. Cancer Res 63:5357-62.
Pregizer S, Barski A, Gersbach CA, Garcia AJ, Frenkel B (2007): Identification of
novel Runx2 targets in osteoblasts: cell type-specific BMP-dependent regulation of
Tram2. J Cell Biochem 102:1458-71.
Prescott J, Jariwala U, Jia L, Cogan JP, Barski A, Pregizer S, Shen HC, Arasheben
A, Neilson JJ, Frenkel B, Coetzee GA (2007): Androgen receptor-mediated
repression of novel target genes. Prostate 67:1371-83.
Qiao M, Shapiro P, Kumar R, Passaniti A (2004): Insulin-like growth factor-1
regulates endogenous RUNX2 activity in endothelial cells through a
phosphatidylinositol 3-kinase/ERK-dependent and Akt-independent signaling
pathway. J Biol Chem 279:42709-18.
Quarles LD, Siddhanti SR, Medda S (1997): Developmental regulation of
osteocalcin expression in MC3T3-E1 osteoblasts: minimal role of the proximal E-
box cis-acting promoter elements. J Cell Biochem 65:11-24.
Reinhold MI, Naski MC (2007): Direct interactions of Runx2 and canonical Wnt
signaling induce FGF18. J Biol Chem 282:3653-63.
Ren B, Robert F, Wyrick JJ, Aparicio O, Jennings EG, Simon I, Zeitlinger J,
Schreiber J, Hannett N, Kanin E, Volkert TL, Wilson CJ, Bell SP, Young RA
(2000): Genome-wide location and function of DNA binding proteins. Science
290:2306-9.
Rice DS, Northcutt GM, Kurschner C (2001): The Lnx family proteins function as
molecular scaffolds for Numb family proteins. Mol Cell Neurosci 18:525-40.
Rittling SR, Matsumoto HN, McKee MD, Nanci A, An XR, Novick KE, Kowalski
AJ, Noda M, Denhardt DT (1998): Mice lacking osteopontin show normal
development and bone structure but display altered osteoclast formation in vitro. J
Bone Miner Res 13:1101-11.
151
Roca H, Phimphilai M, Gopalakrishnan R, Xiao G, Franceschi RT (2005):
Cooperative interactions between RUNX2 and homeodomain protein-binding sites
are critical for the osteoblast-specific expression of the bone sialoprotein gene. J Biol
Chem 280:30845-55.
Roh TY, Ngau WC, Cui K, Landsman D, Zhao K (2004): High-resolution genome-
wide mapping of histone modifications. Nat Biotechnol 22:1013-6.
Salingcarnboriboon R, Tsuji K, Komori T, Nakashima K, Ezura Y, Noda M (2006):
Runx2 is a target of mechanical unloading to alter osteoblastic activity and bone
formation in vivo. Endocrinology 147:2296-305.
Satake M, Nomura S, Yamaguchi-Iwai Y, Takahama Y, Hashimoto Y, Niki M,
Kitamura Y, Ito Y (1995): Expression of the Runt domain-encoding PEBP2 alpha
genes in T cells during thymic development. Mol Cell Biol 15:1662-70.
Sato M, Morii E, Komori T, Kawahata H, Sugimoto M, Terai K, Shimizu H, Yasui
T, Ogihara H, Yasui N, Ochi T, Kitamura Y, Ito Y, Nomura S (1998):
Transcriptional regulation of osteopontin gene in vivo by PEBP2alphaA/CBFA1 and
ETS1 in the skeletal tissues. Oncogene 17:1517-25.
Schroeder TM, Kahler RA, Li X, Westendorf JJ (2004): Histone deacetylase 3
interacts with runx2 to repress the osteocalcin promoter and regulate osteoblast
differentiation. J Biol Chem 279:41998-2007.
Selvamurugan N, Jefcoat SC, Kwok S, Kowalewski R, Tamasi JA, Partridge NC
(2006): Overexpression of Runx2 directed by the matrix metalloproteinase-13
promoter containing the AP-1 and Runx/RD/Cbfa sites alters bone remodeling in
vivo. J Cell Biochem 99:545-57.
Selvamurugan N, Kwok S, Alliston T, Reiss M, Partridge NC (2004): Transforming
growth factor-beta 1 regulation of collagenase-3 expression in osteoblastic cells by
cross-talk between the Smad and MAPK signaling pathways and their components,
Smad2 and Runx2. J Biol Chem 279:19327-34.
152
Selvamurugan N, Pulumati MR, Tyson DR, Partridge NC (2000): Parathyroid
hormone regulation of the rat collagenase-3 promoter by protein kinase A-dependent
transactivation of core binding factor alpha1. J Biol Chem 275:5037-42.
Sevetson B, Taylor S, Pan Y (2004): Cbfa1/RUNX2 directs specific expression of
the sclerosteosis gene (SOST). J Biol Chem 279:13849-58.
Shen J, Hovhannisyan H, Lian JB, Montecino MA, Stein GS, Stein JL, Van Wijnen
AJ (2003): Transcriptional induction of the osteocalcin gene during osteoblast
differentiation involves acetylation of histones h3 and h4. Mol Endocrinol 17:743-
56.
Shen Q, Christakos S (2005): The vitamin D receptor, Runx2, and the Notch
signaling pathway cooperate in the transcriptional regulation of osteopontin. J Biol
Chem 280:40589-98.
Shen R, Chen M, Wang YJ, Kaneki H, Xing L, O'Keefe R J, Chen D (2006): Smad6
interacts with Runx2 and mediates Smad ubiquitin regulatory factor 1-induced
Runx2 degradation. J Biol Chem 281:3569-76.
Shen R, Wang X, Drissi H, Liu F, O'Keefe RJ, Chen D (2006): Cyclin D1-cdk4
induce runx2 ubiquitination and degradation. J Biol Chem 281:16347-53.
Shimoyama A, Wada M, Ikeda F, Hata K, Matsubara T, Nifuji A, Noda M, Amano
K, Yamaguchi A, Nishimura R, Yoneda T (2007): Ihh/Gli2 signaling promotes
osteoblast differentiation by regulating Runx2 expression and function. Mol Biol
Cell 18:2411-8.
Shin CS, Jeon MJ, Yang JY, Her SJ, Kim D, Kim SW, Kim SY (2006):
CCAAT/enhancer-binding protein delta activates the Runx2-mediated transcription
of mouse osteocalcin II promoter. J Mol Endocrinol 36:531-46.
Shirakabe K, Terasawa K, Miyama K, Shibuya H, Nishida E (2001): Regulation of
the activity of the transcription factor Runx2 by two homeobox proteins, Msx2 and
Dlx5. Genes Cells 6:851-6.
153
Shui C, Spelsberg TC, Riggs BL, Khosla S (2003): Changes in Runx2/Cbfa1
expression and activity during osteoblastic differentiation of human bone marrow
stromal cells. J Bone Miner Res 18:213-21.
Sierra J, Villagra A, Paredes R, Cruzat F, Gutierrez S, Javed A, Arriagada G, Olate J,
Imschenetzky M, Van Wijnen AJ, Lian JB, Stein GS, Stein JL, Montecino M (2003):
Regulation of the bone-specific osteocalcin gene by p300 requires Runx2/Cbfa1 and
the vitamin D3 receptor but not p300 intrinsic histone acetyltransferase activity. Mol
Cell Biol 23:3339-51.
Sierra OL, Cheng SL, Loewy AP, Charlton-Kachigian N, Towler DA (2004): MINT,
the Msx2 interacting nuclear matrix target, enhances Runx2-dependent activation of
the osteocalcin fibroblast growth factor response element. J Biol Chem 279:32913-
23.
Smith E, Redman RA, Logg CR, Coetzee GA, Kasahara N, Frenkel B (2000):
Glucocorticoids inhibit developmental stage-specific osteoblast cell cycle.
Dissociation of cyclin A-cyclin-dependent kinase 2 from E2F4-p130 complexes. J
Biol Chem 275:19992-20001.
Sowa H, Kaji H, Hendy GN, Canaff L, Komori T, Sugimoto T, Chihara K (2004):
Menin is required for bone morphogenetic protein 2- and transforming growth factor
beta-regulated osteoblastic differentiation through interaction with Smads and
Runx2. J Biol Chem 279:40267-75.
Spana EP, Doe CQ (1996): Numb antagonizes Notch signaling to specify sibling
neuron cell fates. Neuron 17:21-6.
Stefanovic B, Stefanovic L, Schnabl B, Bataller R, Brenner DA (2004): TRAM2
protein interacts with endoplasmic reticulum Ca2+ pump Serca2b and is necessary
for collagen type I synthesis. Mol Cell Biol 24:1758-68.
Stewart M, Terry A, Hu M, O'Hara M, Blyth K, Baxter E, Cameron E, Onions DE,
Neil JC (1997): Proviral insertions induce the expression of bone-specific isoforms
of PEBP2alphaA (CBFA1): evidence for a new myc collaborating oncogene. Proc
Natl Acad Sci U S A 94:8646-51.
154
Stickens D, Behonick DJ, Ortega N, Heyer B, Hartenstein B, Yu Y, Fosang AJ,
Schorpp-Kistner M, Angel P, Werb Z (2004): Altered endochondral bone
development in matrix metalloproteinase 13-deficient mice. Development 131:5883-
95.
Stock M, Schafer H, Fliegauf M, Otto F (2004): Identification of novel genes of the
bone-specific transcription factor Runx2. J Bone Miner Res 19:959-72.
Stock M, Schafer H, Stricker S, Gross G, Mundlos S, Otto F (2003): Expression of
galectin-3 in skeletal tissues is controlled by Runx2. J Biol Chem 278:17360-7.
Sudhakar S, Li Y, Katz MS, Elango N (2001): Translational regulation is a control
point in RUNX2/Cbfa1 gene expression. Biochem Biophys Res Commun 289:616-
22.
Terry A, Kilbey A, Vaillant F, Stewart M, Jenkins A, Cameron E, Neil JC (2004):
Conservation and expression of an alternative 3' exon of Runx2 encoding a novel
proline-rich C-terminal domain. Gene 336:115-25.
Thirunavukkarasu K, Halladay DL, Miles RR, Yang X, Galvin RJ, Chandrasekhar S,
Martin TJ, Onyia JE (2000): The osteoblast-specific transcription factor Cbfa1
contributes to the expression of osteoprotegerin, a potent inhibitor of osteoclast
differentiation and function. J Biol Chem 275:25163-72.
Thirunavukkarasu K, Mahajan M, McLarren KW, Stifani S, Karsenty G (1998): Two
domains unique to osteoblast-specific transcription factor Osf2/Cbfa1 contribute to
its transactivation function and its inability to heterodimerize with Cbfbeta. Mol Cell
Biol 18:4197-208.
Thomas DM, Carty SA, Piscopo DM, Lee JS, Wang WF, Forrester WC, Hinds PW
(2001): The retinoblastoma protein acts as a transcriptional coactivator required for
osteogenic differentiation. Mol Cell 8:303-16.
Tou L, Quibria N, Alexander JM (2001): Regulation of human cbfa1 gene
transcription in osteoblasts by selective estrogen receptor modulators (SERMs). Mol
Cell Endocrinol 183:71-9.
155
Truong T, Zhang X, Pathmanathan D, Soo C, Ting K (2007): Craniosynostosis-
associated gene nell-1 is regulated by runx2. J Bone Miner Res 22:7-18.
Tsuji K, Ito Y, Noda M (1998): Expression of the PEBP2alphaA/AML3/CBFA1
gene is regulated by BMP4/7 heterodimer and its overexpression suppresses type I
collagen and osteocalcin gene expression in osteoblastic and nonosteoblastic
mesenchymal cells. Bone 22:87-92.
Tsuji K, Komori T, Noda M (2004): Aged mice require full transcription factor,
Runx2/Cbfa1, gene dosage for cancellous bone regeneration after bone marrow
ablation. J Bone Miner Res 19:1481-9.
Vaes BL, Ducy P, Sijbers AM, Hendriks JM, van Someren EP, de Jong NG, van den
Heuvel ER, Olijve W, van Zoelen EJ, Dechering KJ (2006): Microarray analysis on
Runx2-deficient mouse embryos reveals novel Runx2 functions and target genes
during intramembranous and endochondral bone formation. Bone 39:724-38.
van Steensel B, Henikoff S (2000): Identification of in vivo DNA targets of
chromatin proteins using tethered dam methyltransferase. Nat Biotechnol 18:424-8.
Vaughan T, Pasco JA, Kotowicz MA, Nicholson GC, Morrison NA (2002): Alleles
of RUNX2/CBFA1 gene are associated with differences in bone mineral density and
risk of fracture. J Bone Miner Res 17:1527-34.
Vaughan T, Reid DM, Morrison NA, Ralston SH (2004): RUNX2 alleles associated
with BMD in Scottish women; interaction of RUNX2 alleles with menopausal status
and body mass index. Bone 34:1029-36.
Vega RB, Matsuda K, Oh J, Barbosa AC, Yang X, Meadows E, McAnally J,
Pomajzl C, Shelton JM, Richardson JA, Karsenty G, Olson EN (2004): Histone
deacetylase 4 controls chondrocyte hypertrophy during skeletogenesis. Cell 119:555-
66.
Wang C, Lee G, Hsu W, Yeh CH, Ho ML, Wang GJ (2006): Identification of USF2
as a key regulator of Runx2 expression in mouse pluripotent mesenchymal D1 cells.
Mol Cell Biochem 292:79-88.
156
Wang FS, Wang CJ, Sheen-Chen SM, Kuo YR, Chen RF, Yang KD (2002):
Superoxide mediates shock wave induction of ERK-dependent osteogenic
transcription factor (CBFA1) and mesenchymal cell differentiation toward
osteoprogenitors. J Biol Chem 277:10931-7.
Wang Q, Wei X, Zhu T, Zhang M, Shen R, Xing L, O'Keefe RJ, Chen D (2007):
Bone morphogenetic protein 2 activates Smad6 gene transcription through bone-
specific transcription factor Runx2. J Biol Chem 282:10742-8.
Wang S, Wang Q, Crute BE, Melnikova IN, Keller SR, Speck NA (1993): Cloning
and characterization of subunits of the T-cell receptor and murine leukemia virus
enhancer core-binding factor. Mol Cell Biol 13:3324-39.
Wang W, Wang YG, Reginato AM, Glotzer DJ, Fukai N, Plotkina S, Karsenty G,
Olsen BR (2004): Groucho homologue Grg5 interacts with the transcription factor
Runx2-Cbfa1 and modulates its activity during postnatal growth in mice. Dev Biol
270:364-81.
Wee HJ, Huang G, Shigesada K, Ito Y (2002): Serine phosphorylation of RUNX2
with novel potential functions as negative regulatory mechanisms. EMBO Rep
3:967-74.
Wei CL, Wu Q, Vega VB, Chiu KP, Ng P, Zhang T, Shahab A, Yong HC, Fu Y,
Weng Z, Liu J, Zhao XD, Chew JL, Lee YL, Kuznetsov VA, Sung WK, Miller LD,
Lim B, Liu ET, Yu Q, Ng HH, Ruan Y (2006): A global map of p53 transcription-
factor binding sites in the human genome. Cell 124:207-19.
Weinmann AS, Bartley SM, Zhang T, Zhang MQ, Farnham PJ (2001): Use of
chromatin immunoprecipitation to clone novel E2F target promoters. Mol Cell Biol
21:6820-32.
Weinmann AS, Farnham PJ (2002): Identification of unknown target genes of human
transcription factors using chromatin immunoprecipitation. Methods 26:37-47.
Westendorf JJ (2006): Transcriptional co-repressors of Runx2. J Cell Biochem
98:54-64.
157
Westendorf JJ, Zaidi SK, Cascino JE, Kahler R, van Wijnen AJ, Lian JB, Yoshida
M, Stein GS, Li X (2002): Runx2 (Cbfa1, AML-3) interacts with histone deacetylase
6 and represses the p21(CIP1/WAF1) promoter. Mol Cell Biol 22:7982-92.
Wiley SR, Winkles JA (2003): TWEAK, a member of the TNF superfamily, is a
multifunctional cytokine that binds the TweakR/Fn14 receptor. Cytokine Growth
Factor Rev 14:241-9.
Xiao G, Cui Y, Ducy P, Karsenty G, Franceschi RT (1997): Ascorbic acid-dependent
activation of the osteocalcin promoter in MC3T3-E1 preosteoblasts: requirement for
collagen matrix synthesis and the presence of an intact OSE2 sequence. Mol
Endocrinol 11:1103-13.
Xiao G, Jiang D, Ge C, Zhao Z, Lai Y, Boules H, Phimphilai M, Yang X, Karsenty
G, Franceschi RT (2005): Cooperative interactions between activating transcription
factor 4 and Runx2/Cbfa1 stimulate osteoblast-specific osteocalcin gene expression.
J Biol Chem 280:30689-96.
Xiao G, Jiang D, Gopalakrishnan R, Franceschi RT (2002): Fibroblast growth factor
2 induction of the osteocalcin gene requires MAPK activity and phosphorylation of
the osteoblast transcription factor, Cbfa1/Runx2. J Biol Chem 277:36181-7.
Xiao G, Jiang D, Thomas P, Benson MD, Guan K, Karsenty G, Franceschi RT
(2000): MAPK pathways activate and phosphorylate the osteoblast-specific
transcription factor, Cbfa1. J Biol Chem 275:4453-9.
Xiao Z, Awad HA, Liu S, Mahlios J, Zhang S, Guilak F, Mayo MS, Quarles LD
(2005): Selective Runx2-II deficiency leads to low-turnover osteopenia in adult
mice. Dev Biol 283:345-56.
Xiao ZS, Hjelmeland AB, Quarles LD (2004): Selective deficiency of the "bone-
related" Runx2-II unexpectedly preserves osteoblast-mediated skeletogenesis. J Biol
Chem 279:20307-13.
Xiao ZS, Liu SG, Hinson TK, Quarles LD (2001): Characterization of the upstream
mouse Cbfa1/Runx2 promoter. J Cell Biochem 82:647-59.
158
Xiao ZS, Simpson LG, Quarles LD (2003): IRES-dependent translational control of
Cbfa1/Runx2 expression. J Cell Biochem 88:493-505.
Xiao ZS, Thomas R, Hinson TK, Quarles LD (1998): Genomic structure and isoform
expression of the mouse, rat and human Cbfa1/Osf2 transcription factor. Gene
214:187-97.
Yoshida CA, Furuichi T, Fujita T, Fukuyama R, Kanatani N, Kobayashi S, Satake
M, Takada K, Komori T (2002): Core-binding factor beta interacts with Runx2 and
is required for skeletal development. Nat Genet 32:633-8.
Young DW, Hassan MQ, Pratap J, Galindo M, Zaidi SK, Lee SH, Yang X, Xie R,
Javed A, Underwood JM, Furcinitti P, Imbalzano AN, Penman S, Nickerson JA,
Montecino MA, Lian JB, Stein JL, van Wijnen AJ, Stein GS (2007): Mitotic
occupancy and lineage-specific transcriptional control of rRNA genes by Runx2.
Nature 445:442-6.
Zaidi SK, Javed A, Choi JY, van Wijnen AJ, Stein JL, Lian JB, Stein GS (2001): A
specific targeting signal directs Runx2/Cbfa1 to subnuclear domains and contributes
to transactivation of the osteocalcin gene. J Cell Sci 114:3093-102.
Zaidi SK, Javed A, Pratap J, Schroeder TM, J JW, Lian JB, van Wijnen AJ, Stein
GS, Stein JL (2006): Alterations in intranuclear localization of Runx2 affect
biological activity. J Cell Physiol 209:935-42.
Zaidi SK, Sullivan AJ, Medina R, Ito Y, van Wijnen AJ, Stein JL, Lian JB, Stein GS
(2004): Tyrosine phosphorylation controls Runx2-mediated subnuclear targeting of
YAP to repress transcription. Embo J 23:790-9.
Zambotti A, Makhluf H, Shen J, Ducy P (2002): Characterization of an osteoblast-
specific enhancer element in the CBFA1 gene. J Biol Chem 277:41497-506.
Zhang X, Yang M, Lin L, Chen P, Ma KT, Zhou CY, Ao YF (2006): Runx2
overexpression enhances osteoblastic differentiation and mineralization in adipose--
derived stem cells in vitro and in vivo. Calcif Tissue Int 79:169-78.
159
Zhang YW, Yasui N, Ito K, Huang G, Fujii M, Hanai J, Nogami H, Ochi T,
Miyazono K, Ito Y (2000): A RUNX2/PEBP2alpha A/CBFA1 mutation displaying
impaired transactivation and Smad interaction in cleidocranial dysplasia. Proc Natl
Acad Sci U S A 97:10549-54.
Zhao M, Qiao M, Oyajobi BO, Mundy GR, Chen D (2003): E3 ubiquitin ligase
Smurf1 mediates core-binding factor alpha1/Runx2 degradation and plays a specific
role in osteoblast differentiation. J Biol Chem 278:27939-44.
Zhao Z, Zhao M, Xiao G, Franceschi RT (2005): Gene transfer of the Runx2
transcription factor enhances osteogenic activity of bone marrow stromal cells in
vitro and in vivo. Mol Ther 12:247-53.
Zheng H, Guo Z, Ma Q, Jia H, Dang G (2004): Cbfa1/osf2 transduced bone marrow
stromal cells facilitate bone formation in vitro and in vivo. Calcif Tissue Int 74:194-
203.
Zhou G, Zheng Q, Engin F, Munivez E, Chen Y, Sebald E, Krakow D, Lee B (2006):
Dominance of SOX9 function over RUNX2 during skeletogenesis. Proc Natl Acad
Sci U S A 103:19004-9.
Ziros PG, Georgakopoulos T, Habeos I, Basdra EK, Papavassiliou AG (2004):
Growth hormone attenuates the transcriptional activity of Runx2 by facilitating its
physical association with Stat3beta. J Bone Miner Res 19:1892-904.
Ziros PG, Gil AP, Georgakopoulos T, Habeos I, Kletsas D, Basdra EK,
Papavassiliou AG (2002): The bone-specific transcriptional regulator Cbfa1 is a
target of mechanical signals in osteoblastic cells. J Biol Chem 277:23934-41.

Abstract (if available)

Abstract Runx2 is a master transcription factor in osteoblasts, yet its mechanism is poorly understood. In particular, there is a paucity of information about its target genes and their regulation. To address this, we first used ChIP Display to discover novel genomic targets occupied by Runx2 in living MC3T3-E1 osteoblastic cells. One of these targets was located within the promoter of Tram2, whose product facilitates proper folding of type I collagen. We demonstrated that Tram2 mRNA levels were altered by exogenous Runx2, and that this occurred in a BMP- and cell typedependent manner. Thus, Tram2 is likely a Runx2 target gene and may participate in its osteogenic function. Next, we measured endogenous Runx2 expression in MC3T3-E1 cells during development of the osteoblast phenotype, to see if it could explain the dramatic increase in mRNA levels of Osteocalcin (OC), a classic Runx2 target gene. Surprisingly, we discovered that it could not, as Runx2 expression decreased over time, along with in vitro DNA binding activity. Instead, developmental stimulation of OC by Runx2 is attributable to enhanced promoter occupancy in vivo. A remarkably similar pattern of recruitment was observed at the Glt28d2 promoter, a novel Runx2 genomic target discovered by ChIP-Chip analysis of cells in which the OC promoter is maximally occupied. Thus, Runx2 acquires the ability to access target genes relatively late during development of the osteoblast phenotype, and this is most likely due to activation of collaborating factors or to post-translational modification of Runx2 itself. Expanding our knowledge of Runx2 target genes and their regulation is warranted to better understand the regulation of osteoblast function and to provide opportunities for the development of new bone anabolics.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Preosteoblast-specific RUNX2 promotes RANKL membrane association: antagonism by sex steroids through a non-genomic mechanism

PDF

Inhibitory effects of estradiol and SERMs on RUNX2-driven osteoblast differentiation and gene expression

PDF

Glucocorticoids and Runx2 synergistically stimulate Wif1 and compromise pre-osteoblats in vitro

PDF

RUNX2 stimulates osteoblast-driven osteoclastogenesis by RANKL presentation

PDF

RUNX2 & sex steroids: molecular mechanisms in regulating bone turnover

PDF

The role of Runx2 in the development of the tendon-bone attachment unit

PDF

Modulation of Runx proteins by steroid hormone receptors

PDF

The effect of estradiol on TAZ in stromal cell line ST2 in osteoblast differentiation and ER+breast cancer cell

PDF

Study of the role of bone morphogenetic proteins in prostate cancer progression

PDF

Protein arginine methyltransferases in murine skull development

PDF

Systematic analysis of single nucleotide polymorphisms in the human steroid 5-alpha reductase type I gene

PDF

Studies of the biological relevance of Histone H4 Lysine 20 monomethylation: discovery of its role in the cell cycle and localization within the human genome

PDF

DNA hypermethylation: its role in colorectal tumorigenesis and potential clinical applications

PDF

The role of histone H4 lysine 20 monomethylation in gene expression and differentiation

PDF

Mechanisms for regulating the expression of the TATA-binding protein

PDF

Identification of DNA methylation markers in diffuse large B-cell lymphoma

PDF

Epigenetic regulation of non CPG island gene promoters

PDF

Alcohol mediated expression of cyto-protective enzyme - NQO-1 and its post translational regulation

PDF

The role of nuclear GRP78 in regulation of EGFR expression in lung cancer

PDF

The role of Rpd3 and Sir2 in regulation of replication initiation in budding yeast: Rpd3 acts directly on single-copy origins while Sir2 works through ribosomal DNA origins

Asset Metadata

Creator Pregizer, Steven (author)

Core Title Runx2 interactions with the osteoblast genome

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Biochemistry and Molecular Biology

Degree Conferral Date 2008-08

Publication Date 07/15/2008

Defense Date 06/13/2008

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag location analysis,OAI-PMH Harvest,osteoblast differentiation,RUNX2,transcriptional regulation

Language English

Advisor Frenkel, Baruch (committee chair), Coetzee, Gerhard A. (committee member), Hacia, Joseph G. (committee member), Rice, Judd C. (committee member)

Creator Email pregizer@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m1349

Unique identifier UC1431220

Identifier etd-Pregizer-20080715 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-86806 (legacy record id),usctheses-m1349 (legacy record id)

Legacy Identifier etd-Pregizer-20080715.pdf

Dmrecord 86806

Document Type Dissertation

Rights Pregizer, Steven

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu