STEM CELL TRANSCRIPTOMICS:

A SYSTEMS APPROACH TO THE PLURIPOTENT STATE

A DISSERTATION

SUBMITTED TO THE DEPARTMENT OF BIOENGINEERING

AND THE COMMITTEE ON GRADUATE STUDIES

OF STANFORD UNIVERSITY

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

Kitchener Daniel Wilson

May 2010

http://creativecommons.org/licenses/by-nc/3.0/us/

This dissertation is online at: http://purl.stanford.edu/ng707zs0496

© 2010 by Kitchener Daniel Wilson. All Rights Reserved.

Re-distributed by Stanford University under license with the author.

This work is licensed under a Creative Commons Attribution-Noncommercial 3.0 United States License.

ii

I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.

Joseph Wu, Primary Adviser

I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.

Sanjiv Gambhir, Co-Adviser

I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.

Michael Longaker

I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.

Paul Yock

Approved for the Stanford University Committee on Graduate Studies.

Patricia J. Gumport, Vice Provost Graduate Education

This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file inUniversity Archives.

iii

iv

ABSTRACT

The discovery and isolation of human embryonic stem cells (hESCs), and the more recent

generation of induced pluripotent stem cells (iPSCs) from adult cells, has given medical

science the tantalizing prospect of one day regenerating organs and tissues in human patients,

as well as a revolutionary method for investigating heritable human diseases in a petri dish.

This is because hESCs and iPSCs are pluripotent, which enables them to differentiate into

virtually any cell type of the human body. However, forcing these cells to change their

phenotype is an imperfect science, and is often time-consuming, resource-intensive, and

plagued by poor yields. Fundamentally, what is needed is better control over the factors that

induce, maintain, and repress pluripotency. Given the ~25,000 known protein coding genes in

the human genome, as well as numerous known and unknown regulatory elements such as

noncoding RNAs, this is a daunting task. My approach to this problem is taken in large

measure from systems biology, in which understanding biology at the systemic level, rather

than its individual parts, is the central dogma. For example, we now commonly hear words

such as “circuits”, “networks”, and “programs” to describe molecular regulation of cellular

processes. While these terms bring a language of engineering sophistication to biological

research, what they imply is that no single molecule is responsible for a given biological

process. My work over the past four years has therefore attempted to characterize the sets of

molecules, both messenger RNAs and microRNAs, that constitute the “transcriptome” of the

stem cell. This has included understanding how the stem cell transcriptome changes in the face

of external insult (e.g. ionizing radiation), how it changes during differentiation to adult

phenotypes such as cardiomyocytes and endothelial cells, and how knowledge from it may be

used to induce pluripotency. Taken together, interrogating and ultimately controlling the stem

cell transcriptome will be an essential step before we can realize the promise of regenerative

therapy.

v

ACKNOWLEDGMENTS

For their guidance, I am indebted to my mentors Paul Yock, Sanjiv Gambhir, Shan Wang,

Andrew Connolly, Michael Longaker, and especially my research advisor and PI, Joseph Wu,

who has given me invaluable advice – scientifically, professionally and personally -

throughout my graduate studies. Thanks also to the Department of Bioengineering, and in

particular Olgalydia Urbano, for the wonderful graduate program, and to Bio-X for my student

fellowship.

I want to give credit for some of the work presented in this thesis, without which I would not

have a cohesive story to tell. Point-source electrical stimulation of murine embryonic stem

cells was performed by Michael Chen and Xioyan Xie. Zongjin Li performed the

differentiation of endothelial cells from human embryonic stem cells, and acquired the cardiac

imaging data for the rosiglitazone study. Dania Xu performed the experiments for the

microRNA reprogramming efficiency figure (Fig. 25B). Fangjun Jia performed the

experiments to derive iPSCs using the minicircle vector. Zhumur Ghosh performed the

analysis of transcriptional memory in induced pluripotent stem cells (Fig. 27). Thank you all

for your contributions.

In addition to those mentioned above, I owe enormous gratitude to my collaborators and

colleagues in the Wu lab and beyond: Mei Huang, Ning Sun, Andrew Lee, Feng Cao, Shijun

Hu, Oscar Abilez, Jin Yu, Shivkumar Venkatasubrahmanyam (Biomedical Informatics), Atul

Butte (Biomedical Informatics), Robert Robbins, Mark Kay, Todd Brinton, Roger Wagner,

Poornima Parameswaran, John Coller (SFGF), Gergana Nestorova (formerly HIMC), David

Hirschberg (formerly HIMC), Ronald Li (UC Davis), Travis Antes (System Biosciences),

Joan Wilson (formerly System Biosciences), Joseph Gold (Geron), Christoph Eicken (LC

Sciences), and Nichole and Richard Reisdorph (NHLBI Genomics and Proteomics

Workshop).

And last but not least, to my parents, Kitch and Eva, and sisters, Erika and Susanna, for giving

me support and safe haven in Santa Barbara when I needed respite from the cacophony of

Stanford.

vi

TABLE OF CONTENTS

Page Number

LIST OF TABLES vii

LIST OF ILLUSTRATIONS viii

INTRODUCTION 1

1. PERTURBING THE PLURIPOTENT STATE

a. Effects of ionizing radiation on self renewal and

pluripotency of human embryonic stem cells.

5

b. Effects of point source electrical stimulation on mouse

embryonic stem cells.

25

2. DIFFERENTIATING THE PLURIPOTENT STATE

a. Transcriptional profiling of human embryonic stem cell-

derived endothelial cells.

36

b. The miR-145 pluripotency loop in human embryonic stem

cells.

48

c. Dynamic microRNA expression programs during cardiac

differentiation of human embryonic stem cells.

53

3. INDUCING THE PLURIPOTENT STATE

a. MicroRNA profiling of human induced pluripotent stem

cells.

75

b. Current and future directions

i. Improving induction of pluripotency with

microRNAs.

94

ii. Non-viral-derived iPSCs. 99

iii. Comparison of induced pluripotent stem cells that

have been derived from different tissue sources.

103

ADDENDUM

Transcriptome alteration in the diabetic heart by rosiglitazone:

implications for cardiovascular mortality.

106

LIST OF REFERENCES 129

vii

LIST OF TABLES

Page Number

Table 1. Canonical pathways and functions that are disrupted in hESCs at 24

hours after 4 Gy irradiation.

18

Table 2. Gene Set Enrichment Analysis (GSEA) of the expression data for 4

Gy-irradiated hESCs.

19

Table 3. Upregulated Gene Ontology (GO) biological processes in

electrically-stimulated mESCs vs. controls.

29

Table 4. Downregulated GO biological processes in electrically-stimulated

mESCs vs. controls

30

Table 5. Selected canonical pathways predicted to be targets of miR-499. 68

ADDENDUM:

Table 6. Selected genes that are differentially expressed between the hearts of

untreated db/db mice and db/+ controls.

123

Table 7. Selected genes that are differentially expressed between

rosiglitazone-treated and untreated db/db mice.

124

viii

LIST OF ILLUSTRATIONS

Page Number

Figure 1. In vitro studies of irradiated hESCs. 13

Figure 2. Bioluminescence reporter gene imaging of irradiated hESCs in

living animals.

14

Figure 3. In vitro hESC proliferation kinetics after irradiation. 15

Figure 4. Gross image of four subcutaneous teratomas at three radiation

dosages plus control (0 Gy).

15

Figure 5. Microarray analysis of hESCs 24 hours after irradiation 16

Figure 6. Pearson clustering of microarray data from irradiated cells with data

from various differentiated hESC and primary cell types.

21

Figure 7. Image of an assembled microelectrode array chip. 26

Figure 8. Ingenuity Pathways Analysis of transcriptome alteration after

electrical stimulation of mESCs.

31

Figure 9. Transcriptional profiling of human hESC-derived endothelial cells. 38-39

Figure 10. MiRNA biogenesis and targeting. 49

Figure 11. Differentiation of hESCs to cardiomyocytes that express lineage-

specific genes.

62-63

Figure 12. PCA plots of the miRNA-ome data for hESC-CMs. 64

Figure 13. Subsets of miRNAs that are unique to hESCs and hESC-CMs. 65

Figure 14. qPCR confirms the transcriptional changes of selected miRNAs

observed in the microarray data.

66

Figure 15. Target analysis of miR-499, miR-1, and miR-208. 67

Figure 16. Lentiviral vector for miRNA over-expression in hESCs. 70

Figure 17. Over-expression of miR-1 and miR-499 during hESC

differentiation to cardiomyocytes.

71

Figure 18. Electrophysiological changes in hESC-CMs after over-expression

of miR-1 and miR-499

72

Figure 19. Morphology and pluripotent marker staining of human fibroblasts,

iPSCs, and hESCs.

83

ix

LIST OF ILLUSTRATIONS

Page Number

Figure 20. PCR and microarray analysis of human fibroblasts, iPSCs, and

hESCs.

84

Figure 21. PCA plot of the transcriptomes for fibroblasts, iPSCs, and hESCs. 85

Figure 22. Heat maps and signal intensity plots of miRNA expression across

human fibroblasts, iPSCs, and hESCs.

86-87

Figure 23. Quantitative RT-PCR of selected miRNAs confirms expression

patterns seen in the microarray data, and also reveals miRNA expression

during differentiation.

88

Figure 24. Rationale for using microRNAs in combination with transcription

factors to increase reprogramming efficiency.

95

Figure 25. Increased reprogramming efficiencies when using combinations of

miRNAs and transcription factors.

97

Figure 26. Microarray analysis of non-viral-derived iPSCs. 102

Figure 27. Persistent donor cell gene expression among iPSCs contributes to

differences with hESCs.

105

ADDENDUM:

Figure 28. Insulin resistance and mean body/heart weights of rosiglitazone-

treated, untreated, and control mice.

116

Figure 29. High-power light microscopy slides of myocardial tissue from

rosiglitazone-treated and untreated db/db mice, and db/+ controls.

117

Figure 30. Left ventricular functional analysis with echocardiogram of treated

and untreated db/db mice, and db/+ controls.

119

Figure 31. Cardiac magnetic resonance imaging of treated and untreated

db/db mice, and db/+ controls.

120

Figure 32. Multidimensional scaling plot of the microarray data of hearts

from treated and untreated db/db mice, and db/+ controls.

121

1

INTRODUCTION

First isolated by James Thomson and colleagues in 19981, human embryonic stem

cells (hESCs) have inspired both enormous hope and enormous heartache in the public and

scientific communities. When imagining hESC-based therapies of the future, we now think in

terms of actual regeneration rather than just preservation of function: paralyzed patients will

walk again, failing hearts will beat strongly again, weakened pancreas will release insulin

again. These are truly revolutionary ideas, and they speak directly to our relentless search, in

the footsteps of Ponce de León, for the legendary Fountain of Youth. Whereas that

regenerative fountain was based in mythology, we would like to believe that we have the real

thing with stem cells, which are based in science.

After having worked in the field for a handful of years, in my opinion the practical

realization of clinical regenerative medicine remains elusive. Thus far, even in idealized

animal models, our ability to efficiently differentiate stem cells into useful cell types,

transplant them so that they remain alive and integrate with the host parenchyma or stroma,

then ultimately regenerate the diseased area in a meaningful way permanently and consistently

and safely, is still just a dream. There is a mountain of basic science and translational research

that must be scaled before we can routinely inject foreign cells into very sick patients and

expect only positive outcomes that are without risk. That being said, I remain a believer in the

potential of regenerative therapies, and continue to believe that the 21st century will be the

Stem Cell Century. It has been thrilling to spend my graduate years in one of science‟s newest

and fastest-growing fields, made even more thrilling with the invention of a great biological

curiosity, induced pluripotent stem cells (iPSCs).

But first, some background on these cells. The beauty of hESCs is that they are

derived from the inner cell mass of the human blastocyte and, under the right conditions, can

be kept in an undifferentiated, self-renewing state indefinitely. In contrast to adult stem cells,

2

hESCs have the advantage of being pluripotent, which endows them with the ability to

differentiate into virtually every cell type in the human body. However, the use of human

embryos is controversial in the US, and potential tissue rejection following transplantation in

patients remains problematic2. One way to circumvent these issues is to generate iPSCs. In

2006, Shinya Yamanaka was the first to describe induction of pluripotency in mouse

fibroblasts using just four transcription factors (OCT4, SOX2, KLF4, c-MYC)3, and a year

later the derivation of human iPSCs was reported by both Shinya Yamanaka4 and James

Thomson5 (using OCT4, SOX2, NANOG, LIN28). The advantage of this approach is that it

does not require human embryos or oocytes to generate patient-specific stem cells, and

therefore can potentially bypass the ethical and political debates that have surrounded this

field for the past decade. Another important benefit is that for the first time, disease-specific

stem cells can be created, which will help scientists understand the molecular mechanisms of

many common inherited diseases6.

Among the many issues for regenerative medicine, one critical need is a better

understanding of the regulatory mechanisms within stem cells that govern their pluripotency.

This knowledge would lead to better control over the regulatory genes that “pull the strings”

of pluripotency, and allow scientists to more effectively direct their behavior. To do this, an

ideal technology would quickly and accurately analyze the whole genome of the stem cell

under different conditions and states and, with the aid of computers, sift through the enormous

amount of data to uncover important regulatory mechanisms. Thankfully for stem cell

biologists, this “genomics” technology has been developed and perfected since 1977 when

Fred Sanger and colleagues first sequenced the complete genome of the bacteriophage

ΦX1747. Just over two decades later, Sanger‟s advances led to the first sequence of the human

genome in 20018, 9

, representing a landmark in modern biology and new avenues to pursue

global approaches to gene function and its relationship to human physiology.

3

While a major branch of genomics is still concerned with sequencing the genomes of

various organisms, the primary DNA sequence itself can only hint at the molecular function of

any given gene. By necessity, the knowledge of full genomes has spawned the field of

functional genomics, which is primarily concerned with patterns of gene expression under

various conditions. Historically, researchers performed multiple-tissue Northern blots of a

gene‟s expression in a panel of tissues or organs in order to gain a more complete picture of its

function. However, this experiment can be laborious and time-consuming, and the availability

of a representative number of tissue samples can affect the interpretation of results. Many of

these limitations were overcome in 1995 when the use of miniaturized microarrays for gene

expression profiling was first reported by Pat Brown and colleagues10

. The current workhorse

of functional genomics, microarrays allow scientists to construct the equivalent of a multiple-

tissue Northern blot for tens of thousands of genes all at once.

A typical microarray consists of an arrayed series of thousands of spots of DNA

oligonucleotides, each containing picomoles of a specific DNA sequence (“probe”). This can

be a short section of a gene or an oligonucleotide with known specificity that is used to

hybridize a cDNA or cRNA “target”. Probe-target hybridization is usually detected and

quantified by detection of fluorophore- or biotin-labeled targets to determine relative

abundance of nucleic acid sequences in the target. Limitations with microarrays do exist, such

as hybridization and cross-hybridization artifacts, dye-based detection issues, and design

constraints that preclude the detection of different RNA splice patterns and unmapped genes.

For these reasons, next generation sequencing technologies are now supplanting the

microarray for functional genomic studies11-17

. Even with its limitations, microarray

technology has made it possible to rapidly and affordably measure the gene transcript

abundance of whole genomes under any number of cellular environmental conditions, and also

predict the function of poorly understood genes based on their expression profile.

4

Furthermore, because stem cells can be easily perturbed and differentiated in vitro, during

which they exhibit a markedly dynamic range of expression across thousands of genes, they

are an ideal system for in vitro study of the functional genome during human development

and, in the case of iPSCs, the emergence of disease.

The stem cell field is clearly one that requires collaborative, interdisciplinary work –

molecular and cellular biology, developmental biology, genetic engineering, multimodality

imaging, genomics and bioinformatics, clinical medicine, and immunology, to name a few. So

my graduate work has by necessity been diverse in nature, and in this thesis I have included

the majority, but not all, of my published work, with a particular focus on two main fields: (1)

wet lab hESC and iPSC biology, and (2) bioinformatic analysis of the global RNA expression

profiles (the “transcriptomes”) in these cells and their derivatives. This has included

understanding how the stem cell transcriptome changes in the face of external insult (ionizing

radiation18

and electrical pacing19

), how it changes during differentiation to adult phenotypes

such as cardiomyocytes20-22

and endothelial cells23

, and how knowledge from it may be used to

induce pluripotency in adult cells24-27

. In the Addendum, I have included a publication that I

authored in which we used microarrays to understand the mechanism of cardio-toxicity that

has been alleged against the type 2 diabetes drug, rosiglitazone (Avandia)28

. Unfortunately, I

will only discuss some of my work with molecular imaging29, 30

, and will not discuss a recent

publication on nicotinic receptors and hESCs31

.

In this thesis then, I present my complete work on stem cell transcriptomics: a systems

approach to the pluripotent state.

5

1. PERTURBING THE PLURIPOTENT STATE

Effects of ionizing radiation on self renewal and pluripotency of human

embryonic stem cells.

ABSTRACT

Human embryonic stem cells (hESCs) present a novel platform for in vitro investigation of

the early embryonic cellular response to ionizing radiation. Thus far, no study has analyzed

the genome-wide transcriptional response to ionizing radiation in hESCs, nor has any study

assessed their ability to form teratomas, the definitive test of pluripotency. In this study, we

use microarrays to analyze the global gene expression changes in hESCs after low (0.4 Gy),

medium (2 Gy), and high (4 Gy) dose irradiation. We identify genes and pathways at each

radiation dose that are involved in cell death, p53 signaling, cell cycling, cancer, embryonic

and organ development, and others. Using Gene Set Enrichment Analysis (GSEA), we also

show that the expression of a comprehensive set of core embryonic transcription factors is not

altered by radiation at any dose. Transplantation of irradiated hESCs to immune-deficient

mice results in teratoma formation from hESCs irradiated at all doses, definitive proof of

pluripotency. Further, using a bioluminescence imaging technique, we have found that

irradiation causes hESCs to initially die after transplantation, but the surviving cells quickly

recover by two weeks to levels similar to control. To conclude, we demonstrate that similar to

somatic cells, irradiated hESCs suffer significant death and apoptosis after irradiation.

However, they continue to remain pluripotent and are able to form all three embryonic germ

layers. Studies such as this will help define the limits for radiation exposure for pregnant

women and also radiotracer reporter probes for tracking cellular regenerative therapies.

6

INTRODUCTION

Ionizing radiation is a form of electromagnetic radiation produced by x-ray machines,

fluoroscopy, radioactive isotopes, as well as nuclear environmental catastrophe. In pregnant

mothers undergoing diagnostic or therapeutic procedures involving ionizing radiation, or who

may be exposed to environmental radiation, there is a great potential for damage to the early

embryo32, 33

. The current consensus is that exposure to radiation of <0.05 Gy during pregnancy

is not related to an elevated risk of malformation, and many diagnostic procedures remain

below this threshold32, 34

(note that the Gray (Gy) is a unit of absorbed dose and reflects the

amount of energy deposited into a mass of tissue). However, these data are based on limited

human data or on animal models, and so may not accurately reflect the human embryonic

response to radiation.

With the growing number of imaging procedures that employ ionizing radiation such

as x-rays, computed tomographic (CT) scans35-37

, and positron emission tomography (PET) or

single photon emission computed tomography (SPECT) reporter probes that monitor stem cell

transplantation for regenerative and anti-oncogenic therapies38, 39

, as well as concerns over

terrorist attacks involving radioactive materials40

, there is a need to better understand the

effects on human embryonic stem cells (hESCs). A number of reports have studied both UV-

and γ-irradiated mouse41-43

and human44-48

embryonic stem cells, and have primarily focused

on the DNA damage response such as cell cycling, p53 signaling, and apoptosis. Only a two45,

48 have attempted to characterize the effects of radiation on the defining feature of human

embryonic stem cells: pluripotency, or ability to form all three embryonic germ layers.

However, these studies focused on the expression of just two embryonic genes (OCT4,

NANOG) after irradiation, and none have performed teratoma studies to prove pluripotency.

To address this lack of knowledge, we perform gene expression profiling of irradiated

hESCs at three different doses, allowing us to analyze global pluripotency programs that may

7

be affected by radiation. Taking advantage of novel molecular imaging techniques to track

hESC proliferation, we also inject irradiated hESCs into mice and show that despite a transient

decrease in cellular proliferation with the highest dose used in this study (4 Gy), these cells are

still able to ultimately form teratomas. Taken together, we present definitive proof that hESCs

that survive irradiation up to 4 Gy are pluripotent.

MATERIALS AND METHODS

hESC culture. Undifferentiated hESCs (H9 line from Wicell, passages 45 to 55) were grown

on Matrigel-coated plates in mTeSR1 medium (Stem Cell Technologies, Vancouver, BC,

Canada) as previously described49

. Cell media was changed daily, and passaged approximately

every 4-6 days using Collagenase IV. For cell counting, hESC colonies were digested to single

cells with 0.05% trypsin EDTA and counted with a Countess Automated Cell Counter

(Invitrogen).

Irradiation. hESCs were irradiated with 0.4, 2, or 4 Gy of γ-radiation using a Cesium137

irradiator. Immediately after irradiation, cells were returned to the incubator for recovery until

the appropriate time point.

RT-PCR. 18S was used as housekeeping gene control. The primer sets used in the

amplification reaction are as follows:

Human CXCL10 forward primer: 5‟-CTGATTTGCTGCCTTATCTTTCT-3‟

Human CXCL10 reverse primer: 5‟-ACATTTCCTTGCTAACTGCTTTC-3‟

Human GADD45 forward primer: 5‟-TGGAGGAAGTGCTCAGCAAAGCC-3‟

Human GADD45 reverse primer: 5‟-ACGCCTGGATCAGGGTGAAGTGG-3‟

Human 18S forward primer: 5‟-ACACGGACAGGATTGACAGA-3‟

Human 18S reverse primer: 5‟-GGACATCTAAGGGCATCACAG-3‟

8

Immunohistochemical analysis. 48 hours after irradiation, hESCs were fixed with 2%

formaldehyde in PBS for 2 min, permeabilized with 0.5% tritonX-100 in PBS for 10 min, and

blocked with 5% bovine serum albumin in PBS for an hour. Cells were then stained with

appropriate primary antibodies and AlexaFluor conjugated secondary antibodies (Invitrogen).

The primary antibodies for OCT4 (Santa Cruz), SOX2 (Biolegend), NANOG (Santa Cruz),

SSEA-4 (Chemicon), Tra-1-60 (Chemicon), and Tra-1-81 (Chemicon) were used in the

staining.

Annexin V flow cytometry analysis. 48 hours after irradiation, hESCs were harvested and

resuspended in binding buffer and stained with 5 µl of annexin V-fluorescein isothiocyanate

and 5 µl propidium iodide (PI) using the Annexin V : FITC Apoptosis Detection Kit II (cat #

556570, BD Pharmingen). The cell suspension was incubated for 15 min at room temperature

and analyzed on a FACScan flow cytometer (BD Bioscience). Flow cytometry data were

analyzed with FlowJo (Treestar, San Carlos, CA) analysis software.

Microarray hybridization and data acquisition. Total RNA samples were isolated in Trizol

(Invitrogen) followed by purification over a Qiagen RNeasy column (Qiagen) from hESCs 48

hours after irradiation. Three independent experiments for each radiation group plus control

(for a total of 12 unique samples) were harvested for RNA isolation. Using Agilent Low RNA

Input Fluorescent Linear Amplification Kits, cDNA was reverse transcribed from each of 12

RNA samples representing four biological triplicates, as well as the pooled reference control,

and cRNA was then transcribed and fluorescently labeled with Cy5/Cy3. cRNA was purified

using an RNeasy kit (Qiagen, Valencia, CA, USA). 825 ng of Cy3- and Cy5- labeled and

amplified cRNA was hybridized to Agilent 4×44K whole human genome microarrays

(G4112F) and processed according to the manufacturer‟s instructions. The array was scanned

9

using Agilent G2505B DNA microarray scanner. The image files were extracted using Agilent

Feature Extraction software version 9.5.1 applying LOWESS background subtraction and dye-

normalization.

The data were analyzed using GeneSpring GX 10.0 (Agilent Technologies, Santa

Clara, CA) to identify genes which had statistically significantly changed expression between

groups. Genes were considered significantly differentially regulated with P-value < 0.05 and

fold change ≥ 1.4. For hierarchical clustering, we used Pearson correlation for similarity

measure and average linkage clustering. A heat map was generated using Pearson correlation

clustering of a significant gene list after one-way ANOVA of the raw data.

Gene Set Enrichment Analysis (GSEA). GSEA was performed using the GeneSpring GX

software and gene sets downloaded from Molecular Signatures Database (MSigDB) (Broad

Institute, MIT); a custom list of 26 pluripotency genes was also created based on literature

review. Gene sets were considered significant with Q-value <0.25, as recommended50

. Briefly,

the primary result of GSEA is the enrichment score (ES), which reflects the degree to which a

gene set is overrepresented at the top or bottom of a ranked list of genes. The normalized

enrichment score (NES) is the primary statistic for examining gene set enrichment results. By

normalizing the enrichment score, GSEA accounts for differences in gene set size and in

correlations between gene sets and the expression dataset.

Ingenuity Pathway Analysis (IPA). Significant gene lists were generated from the

GeneSpring software and uploaded to IPA for analysis. IPA assigns biological functions to

genes using the Ingenuity Pathways Knowledge Base (Ingenuity Systems, Inc., Redwood

City, CA). This information is used to form networks to create an „interactome‟ of genes that

are involved in specific biological processes.

10

Functional Analysis. The Functional Analysis identified the biological functions

and/or diseases that were most significant to the data set. Molecules from the dataset

that met the P-value cutoff of 0.05 and fold change cutoff of 1.4 were then associated

with biological functions and/or diseases in Ingenuity‟s Knowledge Base were. Right-

tailed Fisher‟s exact test was used to calculate a P-value determining the probability

that each biological function and/or disease assigned to that data set is due to chance

alone.

Canonical Pathway Analysis. Canonical pathways analysis identified the pathways

from the Ingenuity Pathways Analysis library of canonical pathways that were most

significant to the data set. The significance of the association between the data set and

the canonical pathway was measured in two ways: 1) A ratio of the number of

molecules from the data set that map to the pathway divided by the total number of

molecules that map to the canonical pathway is displayed. 2) Fisher‟s exact test was

used to calculate a P-value determining the probability that the association between

the genes in the dataset and the canonical pathway is explained by chance alone.

Generation of stable reporter gene hESC lines. Enhanced green fluorescent protein (eGFP)

and firefly luciferase (Fluc) double fusion reporter gene positive hESCs (Fluc+/eGFP+

hESC) have

been previously described 20, 29

. Briefly, SIN lentivirus was prepared by transient transfection

of 293T cells. H9 hESCs were stably transduced with LV-pUB-Fluc-eGFP at a multiplicity of

infection (MOI) of 10. The infectivity was determined by eGFP expression as analyzed on a

FACScan. eGFP positive cell populations were isolated by fluorescence activated cell sorting

(FACS) Vantage SE cell sorter (Becton Dickinson), followed by plating for long-term culture.

11

Subcutaneous transplantation of hESCs. Animal protocols were approved by the Stanford

University Animal Care and Use Committee guidelines. All procedures were performed on 8-

10 week old female SCID Beige mice (Charles River Laboratories, Wilmington, MA).

Following induction with inhaled isoflurane (2% to 3%), anesthesia was then maintained with

1% to 2.5% isoflurane. 200,000 Fluc+/eGFP+

hESCs were suspended in a 50 µl volume of a 1:1

mixture of growth factor reduced-Matrigel and DMEM, then irradiated at the appropriate

dosage (0.4, 2 or 4 Gy). Irradiated cell suspensions were each injected subcutaneously into the

dorsum of eight SCID mice; injections were performed within 2 hours of irradiation.

Bioluminescence imaging of transplanted cell survival. Bioluminescence imaging was

performed using the Xenogen IVIS 200 system (Caliper Life Sciences, Hopkinton, MA). After

intraperitoneal injection of the reporter probe D-Luciferin (375 mg Luciferin/kg body weight),

animals were imaged for 1-10 minutes. The same mice were imaged for 6 weeks. Regions of

interest (ROI) were drawn over the signals using the Igor image analysis software

(Wavemetrics, Lake Oswego, OR). BLI signal was standardized for acquisition time and

quantified in units of maximum photons per second per square centimeter per steridian

(photons/sec/cm2/sr).

Postmortem immunohistochemical staining. Animals were sacrificed according to protocols

approved by the Stanford Animal Research Committee after the duration of the study.

Teratomas were explanted and processed for H&E staining. Slides were interpreted by an

expert pathologist.

12

Statistical analysis. Non-microarray data are presented as mean S.D. Data were compared

using standard or repeated measures, using ANOVA where appropriate. Differences were

considered significant for P<0.05.

RESULTS

Fig. 1A gives a schematic of our experimental design. At the high dose of 4 Gy, we

observed massive cell death that was concurrent with the development of holes and patchy

regions in hESC colonies at 48 hours (Fig. 1B,C); hole formation has also been reported in

colonies six hours after 5 Gy irradiation45

. However, the surviving hESCs continued to express

common pluripotency markers such as TRA-1-81, SSEA4, TRA-1-60, and embryonic

transcription factors such as OCT4, SOX2, and NANOG that are key regulators of

pluripotency and self-renewal (Fig 1C).

We were curious about the relative extent of apoptosis and cell death after irradiation

at the different dosages, and so double-stained hESCs with Annexin V (early apoptosis) and

PI (cell death) 48 hours after irradiation, and analyzed the cells with flow cytometry (Fig. 1D).

Clearly, there is a trend towards increasing apoptosis and cell death at the higher radiation

doses (2 and 4 Gy) compared to low dose (0.4 Gy) and control. The majority (>70%) of

hESCs are dead after 4 Gy irradiation, though an apoptotic minority (<30%) appears to

survive at 48 hrs. This latter population likely represents the surviving cells that continue to

express pluripotency markers, as seen in Fig. 1C. However, the definitive test of pluripotency

of human cells is the ability to form a teratoma, which we performed next.

To confirm that surviving hESCs are pluripotent, we injected irradiated cells into

immuno-compromised mice and monitored for teratoma formation. We tracked their growth

kinetics in vivo by using hESCs that constitutively express a Fluc-eGFP double fusion reporter

13

Figure 1. In vitro studies of irradiated hESCs. (A) Schematic of experimental setup. (B,C)

Immunostaining of pluripotency markers in control (B) and irradiated (C) hESCs shows

maintenance of marker expression 48 hours after irradiation. (D) Flow cytometry of FITC

Annexin V and propidium iodide (PI) double-stained hESCs 48 hours after the indicated

radiation dose (one representative experiment of three is shown).

14

gene (Fig. 2A), enabling longitudinal monitoring of cellular photon emission, and by

extension their proliferation, as described20, 29

. After irradiation and injection of

Fluc+/eGFP+hESCs, we found that photon emission from the 2 and 4 Gy groups reached a

statistical minimum at seven days that was less than photon emission from cells exposed to 0

and 0.4 Gy, suggesting massive cell death (P<0.05, n=8 per group, see Fig. 2B,C). Based on

the photon intensities, we estimated that 38±30%, 63±20%, 80±9%, and 83±7% (mean±SEM)

of the 0, 0.4, 2, and 4 Gy-irradiated cells, respectively, had died at day 7. We expected that the

Figure 2. Bioluminescence reporter gene imaging of irradiated hESCs in living animals.

(A) The double fusion reporter gene construct carrying firefly luciferase (Fluc) and enhanced

green fluorescent protein (eGFP). (B) Diagram of the subcutaneous injection sites for each

radiation group plus control, as well as representative bioluminescent images for three mice

through day 42. (C) Plot of longitudinal bioluminescent signal intensities for each group (*0

and 0.4 Gy vs. 2 and 4 Gy, P<0.05). Data presented as mean±SEM. (D) H&E staining of

teratoma section from representative 4 Gy group demonstrating three embryonic germ layers.

15

2 and 4 Gy groups would continue to die,

but surprisingly all four groups emitted

similar levels of photons by day 21,

indicating that surviving hESCs had

recovered from high dose irradiation. We

confirmed this result by studying long-term

in vitro cultures of irradiated hESCs, and

found that cell proliferation was inhibited in

the first week after high dose irradiation, but

thereafter all groups exhibited similar

growth kinetics (Fig. 3). Note that after the post-irradiation “recovery period”, we did not

observe any compensatory increase in cell

proliferation in the high dose groups. Finally, of

the eight mice used in this study, five

developed teratomas in the 4 Gy group by the

sixth week (see Fig. 2D for representative H&E

images, and Fig. 4 for a representative gross

image of four teratomas from a single mouse).

The three mice that failed to form teratomas in

the 4 Gy group likely experienced significant

apoptosis and cell death, and not loss of

pluripotency. To confirm this, we performed a

careful microarray study of the core set of

pluripotency genes to determine whether there

are any detectable changes in pluripotency

Figure 4. Gross image of four

subcutaneous teratomas at three

radiation dosages plus control (0 Gy).

Each teratoma has its own blood vessel

network, demonstrating nascent nutrient

and oxygen supply. Mouse was sacrificed

at 6 weeks after subcutaneous injection of

irradiated hESCs.

Figure 3. In vitro hESC proliferation

kinetics after irradiation. Cells were

irradiated on Day 0. The fold change in total

cell number for each group was calculated by

counting the cells at the beginning and end of

the indicated range of days. Experiments were

performed in duplicate.

16

programs, however subtle, in response to ionizing radiation.

For the transcriptomic analysis of irradiated hESCs, RNA was isolated from cells 24

hours after irradiation, then labeled and hybridized to microarrays (raw data files have been

uploaded to GEO under accession number GSE20951). When analyzing microarray data, it is

often informative to start from a system-wide rather than individual-gene view of the resulting

data, especially when the overall gene fold changes are no more than seven-fold. An overview

of the gene profiles can be seen in the heat map of Fig. 5A. Most apparent is the co-clustering

of the control and low dose samples (0 and 0.4 Gy), which were distinct from the co-clustering

of the high dose samples (2 and 4 Gy). This pattern is also evident in Fig. 5B, in which global

Figure 5. Microarray analysis of hESCs 24 hours after irradiation. (A) Pearson

clustering of the data for 0, 0.4, 2, and 4 Gy irradiated hESCs (n=3 biological replicates

per group). Note that one replicate from the 0.4 Gy group was lost due to poor array

hybridization. Each gene is represented by a single row and each sample by a single

column. Red = upregulated, Green = downregulated. (B) Global Pearson correlation of

microarray data. (C) Venn diagram of the significant entities (P<0.05, fold change ≥1.4)

between each radiation group and control (0 Gy).

17

Pearson correlation shows 95% correlation between the 2 and 4 Gy groups, but only 86%

correlation with the low-dose 0.4 Gy group. A Venn diagram of the entities that are

significantly different (P<0.05, fold change ≥1.4) between radiation dosage and control further

illustrates this pattern (Fig. 5C, note that microarrays often contain multiple probes, or

“entities”, for a given gene). Again, we observe the same grouping as in the Pearson

correlation: the 2 and 4 Gy-irradiated samples exhibit a higher degree of overlap between

themselves than they do with the 0.4 Gy group.

We next used Ingenuity Pathways Analysis (Ingenuity® Systems,

www.ingenuity.com) and Gene Set Enrichment Analysis (GSEA)50

to further analyze the

microarray data. Selected canonical pathways and functions that are disrupted after 4 Gy

irradiation (vs. control) are summarized in Table 1. After 4 Gy irradiation, canonical

pathways such as VDR/RXR activation, p53 signaling, aryl hydrocarbon signaling, and

functions such as cancer, cell death, cell cycle, growth and proliferation, and embryonic

development are significantly affected in hESCs. Specifically, several tumor protein p53

associated genes such as TP53Inp1 (up 2.6-fold) and target genes such as CDKN1A (up 2-

fold) and MDM2 (up 1.7-fold)51

, as well as several tumor necrosis factor receptor superfamily

members, were disregulated after irradiation. A small group of genes associated with

development also exhibited differential expression, including HES1 (down 1.8-fold)52

, Runx1

(up 1.5 fold), and PBX1 (down 1.8-fold); note that many of these genes are also associated

with cancer (Table 1). Supporting the observation that genes related to cancer are disregulated

with radiation, GSEA, a method for analyzing a priori gene sets within microarray data,

revealed upregulation of gene sets that have also been reported in cells after treatment with

chemotherapeutic drugs53-55

(Table 2).

We have also analyzed the progression of gene and pathway changes that occur in

hESCs at each increasing radiation dose: between 0 and 0.4 Gy, 0.4 and 2 Gy, and 2 and 4 Gy.

18

Table 1. Canonical pathways and functions that are disrupted in hESCs at 24 hours after

4 Gy irradiation. Determined with Ingenuity Pathway Analysis.

CANONICAL PATHWAYS GENES

VDR/RXR Activation IGFBP6, CDKN1A, CSNK2A1, HES1, RXRB, PRKCB

p53 Signaling TP53INP1, CDKN1A, TNFRSF10B, C12ORF5, MDM2, HIPK2

Aryl Hydrocarbon Receptor

Signaling TFF1, NQO1, CDKN1A, MDM2, DHFR, RXRB, AHR

PDGF Signaling CSNK2A1, INPP5D, PRKCB

NOTCH Signaling DTX1, HES1

Cell Cycle: G2/M DNA

Damage: Checkpoint Regulation CDKN1A, MDM2

Molecular Mechanisms of

Cancer HHAT, RALA, CDKN1A, MDM2, HIPK2, GLI1, FAS, PRKCB

FUNCTIONS GENES

Cancer

DPYD, PLK3, PBX1, EIF4A2, ATP4A, HES1, DDB2, LATS2, TTC22, FHL2, TFF1,

MPHOSPH8, NEK6, CSNK2A1, TNFRSF10C, CBFB, TUBA1C, TTC5, HIPK2,

CDKN1C, AHR, MYO6, IGFBP6, IFIT3, TUBB3, TP53INP1, NKTR, PPP1R1B, TNFRSF10B, L1CAM, UBE2S, RASD1, CD40, MIB1, LRP8, RUNX1, DHFR, HELB,

EBI3, RPL13A, KIAA1370, POLE3, FBXL7, PIAS1, GDF15, FAS, STK17A, GBP2,

RXRB, VASP, CALR, ATP1B1, CKM, NQO1, TOB1, MDM2, SLC7A8, FDXR, INPP5D, DNM1, TRPM6, HP, GMPS, CDKN1A, MEF2C, GLI1, COBLL1, POLH,

PRKCB

Cell Death

QKI, PLK3, PBX1, DDB2, HES1, TREM2, LATS2, FHL2, BLOC1S2, TFF1, MTF1, NEK6, CSNK2A1, CBFB, TTC5, TNFRSF10C, CDKN1C, HIPK2, AHR, MYO6,

SHISA5, IGFBP6, TUBB3, TP53INP1, SGCG, PPP1R1B, TNFRSF10B, L1CAM,

UBE2S, LAX1, RASD1, NCF1, MNT, CD40, MIB1, RUNX1, GDF15, PIAS1, FAS, STK17A, RPS3, SOX5, CALR, THG1L, NQO1, CDC42EP3, MDM2, INPP5D, FDXR,

DNM1, CDKN1A, MEF2C, GLI1, PRKCB, POLH

Cell Morphology RALA, PLK3, GDF15, ATP4A, HES1, LATS2, FAS, ANK1, MACF1, CSNK2A1,

CBFB, CDKN1C, RXRB, VASP, AHR, PRICKLE2, TNFRSF10B, MDM2, L1CAM, DNM1, PIP5K1A, CD40, MIB1, CDKN1A, FOXJ1, B3GAT1, LRP8, NDST2, GLI1

Cellular Assembly and

Organization RALA, TUBGCP3, NQO1, TNFRSF10B, PLEC1, MDM2, FAS, ANK1, DNM1,

TUBGCP5, RAB11FIP4, AMPH, CDKN1A, TTL, NDST2, CDKN1C, VASP, AHR

Embryonic Development TP53INP1, PIAS1, TNFRSF10B, PBX1, L1CAM, HES1, FAS, DNM1, MTF1,

RUNX1, HIPK2, GLI1, VASP, AHR

Cellular Growth and

Proliferation

GDF15, PIAS1, PBX1, HES1, FAS, FHL2, HS6ST2, CBFB, CDKN1C, HIPK2,

VASP, AHR, CALR, IGFBP6, TUBB3, TP53INP1, TNFRSF10B, TOB1, PTP4A1,

MDM2, INPP5D, TRPM6, NCF1, MNT, CD40, CDKN1A, UHMK1, DHFR, RUNX1, GLI1, PRKCB

Cell Cycle CALR, TP53INP1, GDF15, PIAS1, MDM2, HES1, LATS2, INPP5D, FAS, DNM1,

SESN1, MNT, CD40, CDKN1A, NEK6, CSNK2A1, UHMK1, RUNX1, CDKN1C, HIPK2, AHR

Cellular Development

QKI, PBX1, HES1, TREM2, FAS, ANK1, FOXN4, FHL2, DTX1, TFF1, CBFB,

CDKN1C, RXRB, AHR, VASP, SOX5 (includes EG:6660), MYO6, IGFBP6, CALR,

CTSK, TUBB3, PPP1R1B, DNER, TNFRSF10B, TOB1, L1CAM, MDM2, INPP5D, NCF1, MNT, CD40, MIB1, CDKN1A, MEF2C, RUNX1, GLI1, PRKCB

19

Similar to 4 Gy irradiation, 0.4 Gy irradiation affects cellular functions such as cell death,

cancer, and signaling pathways such as p53, though not important p53 downstream target

genes such as CDKN1A and MDM2. Because CDKN1A is an important negative regulator of

cell cycling56

, the lack of upregulation of CDKN1A by 0.4 Gy irradiation could partly explain

why we did not observe a similar reduction in cell proliferation as in the 2 and 4 Gy groups.

Table 2. Gene Set Enrichment Analysis (GSEA) of the expression data for 4 Gy-

irradiated hESCs. ES=Enrichment Score, NES=Normalized Enrichment Score.

GENE SET

NAME

P-

value

Q-

value ES NES DESCRIPTION GENES IN SET

METHOTREXATE_

PROBCELL_UP <0.001 0.1447 0.6866 1.8669

Upregulated in pro-

B cells (FL5.12)

following treatment

with methotrexate

(Brachat et al.)

ABI1, ABLIM1, CARHSP1, CASP4, CDKN1A, EI24,

H2AFJ, LPIN1, LRRC2,

MALAT1, MDM2, PVRL4, SLC7A14, TOB1, TP53INP1,

TRAFD1, TXNIP, UCHL5

BLEO_HUMAN_LY

MPH_HIGH_4HRS_

UP <0.001 0.1929 0.7732 1.8725

Upregulated at 4

hours following

treatment of human

lymphocytes (TK6)

with a high dose of

bleomycin (Islaih et

al.)

BBC3, BTG1, BTG2,

CDKN1A, CLK1, DDB2,

DDIT4, DUSP14, ENC1, FAS,

FDXR, GADD45A, GDF15,

HNRPA1, IER3, PLXNB2, PMAIP1, PPM1D,

TNFRSF10B, TNFSF9, XPC

CAMPTOTHECIN_P

ROBCELL_UP <0.001 0.2290 0.6617 1.8162

Upregulated in pro-

B cells (FL5.12)

following treatment

with camptothecin

(Brachat et al.)

ABI1, C12ORF22, C1R,

CARHSP1, CBS, CDKN1A, EI24, GRB10, H2AFJ, LPIN1,

LRRC2, NUDCD2, PMM1,

PVRL4, SERTAD1, SLC7A14, TOB1, TP53INP1, TPP1,

TRAFD1, TXNIP, UCHL5, ULK1

OXSTRESS_BREAS

TCA_UP <0.001 0.2314 0.7095 1.8060

Upregulated by

H2O2, Menadione

and t-BH in breast

cancer cells

(Chuang et al.)

C1ORF107, CTAGE5,

CYP1B1, DDIT3, DKK1,

EDN1, EGR1, EPS8L2, FAS, FDXR, GDF15, H19, HMOX1,

HSPA1B, JUN, LIF. LRP6,

MT1H, MT1X, NQO1, PLA2R1, PPP2CB,

PSITPTE22, PSMD12,

PSMD3, RCBTB1, RPL38, SLC35B3, ZBTB4

Human Embryonic

Stem Cell Markers 0.4000 0.5000 -0.3177 -1.0009

Common human

embryonic stem cell

genes (based on

literature search)

POU5F1, NANOG, KLF2,

KLF5, SOX1, SOX2, SOX5, LIN28, DNMT3B, GDF3,

DPPA4, DPPA5, ESRRG,

SALL4, NR6A1, TDGF1, TBX3, FOXD3, FGF4, ZFP42,

LEFTY1, LEFTY2, ERAS,

PODXL, TERT, UTF1

20

Relative to 0.4 Gy irradiation, 2 Gy irradiation affects canonical TFG-β and WNT/β-catenin

signaling, including the genes TGFBR2 (up 1.4-fold), WNT1 (up 1.4-fold), WNT10A (up 2.1-

fold) and WNT9A (up 1.8-fold); notably, WNT proteins play important and diverse roles in

embryonic stem cells57

. 2 Gy irradiation also induces CDKN1A upregulation by 2.3-fold, but

not MDM2. Interestingly, many genes involved in functions such as cellular compromise,

amino acid metabolism, molecular transport, and cell morphology, in addition to cancer and

cell death, were significantly disrupted by 2 Gy of radiation, including a number of solute

carrier family proteins such as SLC6A13 (up 2-fold) and SLC25A13 (down 2.2-fold). Clearly

there is an overall significant increase in cellular dysfunction after 2 Gy irradiation, which

explains the results from our in vivo and in vitro studies (Figs. 1 and 2). Finally, in the 4 vs. 2

Gy group, the overall gene changes were not large, but a small group of genes related to organ

and tissue development did have altered expression, such as TNFSF11 (up 1.6-fold), OTX1

(down 1.6-fold), B4GALT1 (down 1.4-fold), and MEF2C (up 1.9-fold). Presumably there are

subtle development and differentiation processes that are activated with 4 Gy irradiation, but

are not robust enough to cause loss of pluripotency, as evidenced by successful formation of

teratomas from 4 Gy-irradiated hESCs. Furthermore, the gene expression profiles of irradiated

cells showed no increased correlation with previously reported data for differentiated hESCs

and primary cell types20, 23, 26

, giving additional evidence that differentiation is not

significantly increased with ionizing radiation (Fig. 6).

With these observations, we were particularly interested to understand whether core

pluripotency genes were also affected by ionizing radiation. Importantly, well-known

embryonic transcription factors such as OCT4 (POU5F1), SOX2, and NANOG, which are

expressed exclusively or predominantly in hESCs and are critical for maintaining pluripotency

and self-renewal, were not present in any of our significant gene lists across all radiation

dosages. Moving beyond individual genes, we created a gene set of 26 known factors that are

21

well-known to be specific to hESCs (bottom row of Table 2). GSEA revealed no significant

up- or down-regulation of this set in the microarray data for the highest radiation dose, 4 Gy

(P=0.4, Q=0.5, NES=-1.0). This finding agrees with a previous report that showed

normalization of OCT4 and NANOG after only 24 hours in 2 Gy-irradiated hESCs45

. We

therefore conclude that pluripotency gene programs are not significantly affected by high dose

radiation, and this accounts for the observation that surviving hESCs are still capable of

forming all three embryonic germ layers.

Figure 6. Pearson clustering of microarray data from irradiated cells with data from

various differentiated hESC and primary cell types. Microarray data was selected from

samples submitted to GEO (GSE20951, GSE13834, GSE20013, and GSE20014). hESC-CM

= cardiomyocytes derived from H9 human embryonic stem cells; hESC-EC = endothelial

cells derived from H9 human embryonic stem cells; HUVEC = human umbilical vein

endothelial cells; IMR90 = human foreskin fibroblasts; hASC = adult human adipose stem

cells. Each sample depicted in the dendrogram represents the average of multiple biological

replicate data.

22

DISCUSSION

In pregnant mothers undergoing diagnostic or therapeutic procedures involving

ionizing radiation, or who may be exposed to environmental radiation, there is a great

potential for damage to the early embryo. Although the embryo is somewhat protected by the

uterus, it is particularly sensitive to ionizing radiation, and the developmental consequences

can be quite serious33

. Data regarding the potential biological effects on the embryo after in

utero irradiation are based on the results of animal studies58-61

and a limited number of human

exposures such as the 1945 atomic bomb survivors from Hiroshima and Nagasaki. Based on

these collective data, it has been well established that the dose of ionizing radiation, and the

developmental stage of the embryo, are the determining factors for reproductive toxicity in

embryonic development32

. Above poorly-defined threshold doses, the major effects of ionizing

radiation are lethality during the preimplantation–preorganogenetic period, and malformations

and growth retardation during organogenesis32, 34

. Other sequelae later in life may include

severe mental retardation, a reduced intelligence quotient, and childhood cancer32, 34, 62

.

Unfortunately the absolute incidence and radiation dose at which these changes occur, as well

as the mechanisms of damage, remain unclear.

Because the in utero embryonic response to ionizing radiation is not well understood

due to the obvious ethical concerns of exposing pregnant mothers to radiation, hESCs present

a novel in vitro platform for studying the human embryonic response to irradiation. These

cells are derived from the inner cell mass of the blastocyst during embryonic development,

and are therefore closely related to the early stage human embryo. Admittedly, hESCs are still

different from the early embryo in that they lack the complex and dynamic signaling

environment of the uterus and are instead maintained long-term in relatively simple in vitro

cultures. However, the advantage is that we can begin to tease out the embryonic response to

irradiation in a human rather than murine system. Furthermore, because radiotracers and PET

23

reporter genes that monitor cellular transplantation for emerging regenerative and anti-

oncogenic therapies are being increasingly employed in laboratory research39

, and one day

may even achieve routine clinical application38

, it will be important to determine whether such

radioactive probes can directly affect the viability and function of the transplanted cells. For

these reasons, we decided that a broad survey of the functional and global molecular response

of hESCs to irradiation, and in particular the radiation‟s effect on pluripotency, was a critical

area of investigation.

Not surprisingly, our results show that high doses of radiation cause massive cell

death, with a trend towards increasing apoptosis and death at the higher radiation doses (2 and

4 Gy) compared to low dose (0.4 Gy) and control. Interestingly, an apoptotic minority does

appear to survive at 48 hours. Using a bioluminescence imaging technique, we confirmed that

the higher doses of radiation cause hESCs to initially die after transplantation, but the

surviving cells recover by two weeks to levels similar to control. Regardless of the radiation

dose used in our study, all groups of irradiated hESCs were able to form teratomas, the

definitive test of pluripotency. Our genome-wide analysis of gene expression revealed genes

and pathways at each radiation dose that are involved in cell death, p53 signaling, cell cycling,

cancer, embryonic and organ development, and others. Importantly, GSEA showed that the

expression of a comprehensive set of core embryonic transcription factors is not significantly

altered by radiation at any dose, and helps explain how irradiated hESCs are still able to form

teratomas.

In summary, this is the first study of hESC genome-wide transcriptional changes

induced by ionizing radiation, and is a preliminary step towards a better understanding of not

only hESC molecular changes but also the in utero embryonic gene expression response. We

have shown that, similar to somatic cells, irradiated hESCs suffer significant death and

apoptosis after irradiation. Though some gene programs involved in developmental pathways

24

are altered with high dose radiation, the expression of pluripotency genes is unaffected, and

these cells can still form teratomas. Studies such as this may help define the limits for

radiation exposure for pregnant women and also radiotracer reporter probes for tracking

cellular regenerative therapies.

25

1. PERTURBING THE PLURIPOTENT STATE

Effects of point source electrical stimulation of mouse embryonic stem cells19

.

INTRODUCTION

Stem cell therapy is emerging as a promising clinical approach for myocardial repair.

However, the interactions between the graft and host, resulting in inconsistent levels of

integration, remain largely unknown. In particular, the influence of electrical activity of the

surrounding host tissue on graft differentiation and integration is poorly understood. In order

to study this influence under controlled conditions, an in vitro system was developed.

Electrical pacing of differentiating murine embryonic stem cells (mESCs) was performed at

physiologically relevant levels through direct contact with microelectrodes, simulating the

local activation resulting from contact with surrounding electroactive tissue. Among other

cellular assays, cells stimulated with electrical activity up to four days were analyzed by

genome microarray analysis, revealing broad transcriptome changes after pacing. Concurrent

to upregulation of mature gene programs including cardiovascular, neurological, and

musculoskeletal systems is the apparent downregulation of important self-renewal and

pluripotency genes. Overall, a robust system capable of long-term stimulation of mESCs is

demonstrated, and specific conditions are outlined that encourage cardiomyocyte

differentiation.

MATERIALS AND METHODS

Electrical Stimulation. The microelectrode arrays (MEAs) used for this study contain

stimulation electrodes symmetrically arranged across the surface with varying surface

geometries (Fig. 7), and have been previously described63

. All of the outer stimulation

electrodes were connected in parallel and then used to stimulate the cells with the same signal.

26

Cell Culture. Murine ES-D3 cell line

(CRL-1934) was obtained from the

American Type Culture Collection

(ATCC; Manassas, VA). To develop the

mESCs into embryoid bodies (EBs), the

“hanging drop” method was used. LIF

was withdrawn from the medium, and a

cultivation of about 400 cells was

suspended in 18 μL hanging drops to

form EB. At this point, the

differentiation stage of the EB was noted

as Day 0. After 2 days, each EB was

transferred into its own well in an ultra-

low attachment 96-well plate (Corning

Life Sciences, Lowell, MA) for 2 days,

after which they were further seeded onto 48-well plates. At the desired differentiation stage,

whole EBs were dissociated with Collagenase B (Worthington, Lakewood, NJ) and cells were

plated onto the MEA surface at a density of 32,000 cells/cm2 (80 K total cells). At least 2 h

prior to plating, the MEA surface was coated with hESC-qualified matrix Matrigel (BD,

Franklin Lakes, NJ). The cells were allowed to settle between 20 and 30 h before electrical

simulation was applied for 4 days.

Genome Microarray: Labeling Reaction, Hybridization, and Data Acquisition. All

sample processing for microarray analysis was performed at the same time to negate potential

technical variability. Using Low RNA Input Fluorescent Linear Amplification Kits (Agilent

Figure 7. Image of an assembled

microelectrode array chip. Chip is glued and

wire-bonded to a printed circuit board carrier.

An array of 6×6 electrodes located in the center

of the chip is used for electrical sensing. The

larger electrodes on the periphery (contained in

the dashed boxes) are used for stimulation19

.

27

Technologies, Santa Clara, CA), cDNA was reverse transcribed from each RNA sample (N =

8 per group), and cRNA was then transcribed and fluorescently labeled from each cDNA

sample. cRNA samples derived from four biological replicates of a pooled collection of

electrically stimulated ESCs, as well as a pooled reference of RNA taken from control (non-

stimulated) ESCs, were labeled with Cy5 and Cy3, respectively. A mixture of 825 ng of Cy3-

and Cy5-labeled and amplified cRNA was fragmented according to the Agilent technology

protocol. cRNA was hybridized to 4×44K whole mouse genome microarray slides and

scanned according to Agilent‟s instructions. The image files were extracted using Agilent

Feature Extraction software version 9.5.1 applying LOWESS background subtraction and dye

normalization.

Genome Microarray: Data Analysis. Microarray data analysis was performed using

GeneSpring GX 7.3.1 (Agilent Technologies, Santa Clara, CA). Genes were considered

significant if the fold change was greater or less than 1.4 and had a P-value of less than 0.05.

Because of the small changes in overall transcription between the two groups, multiple testing

corrections to remove false positives in the resulting data were not possible. Further analysis

was performed using GeneSpring‟s Gene Ontology browser and Ingenuity Pathway Analysis

(Ingenuity Systems, Redwood City, CA).

RESULTS AND DISCUSSION

Global gene expression profiling of embryonic stem cells enables a systems-based

analysis of the biological processes, networks, and genes that drive cell fate decisions. To

understand these transcriptome changes, a genome microarray analysis of electrically

stimulated murine EBs at 30 μA versus non-stimulated controls was performed. Using a 1.4-

fold change cutoff and P<0.05, we observed 495 upregulated genes and 729 downregulated

28

genes in EBs that had been electrically stimulated. Overall, the fold change in expression level

of these genes was less than 3.5 compared to controls, so on average electrical stimulation did

not dramatically alter the transcriptome. However, a number of interesting developmental-

related genes exhibited altered expression. Perhaps most significant is the downregulation of

OCT4 (POU5F1) and FOXD3 in electrically-stimulated EBs, with a corresponding

upregulation of important differentiation and developmental-related genes such as NCAM1,

ISL1, FOXC1 and FOXC2. We will discuss these genes and more below.

The apparent alteration in developmental gene programs of electrically stimulated

EBs is clearly seen with Gene Ontology (GO) overrepresentation analysis, which categorizes

genes based on their annotations into functional groups (Tables 3 and 4). The upregulated GO

terms in stimulated EBs reveals a number of processes related to embryonic development,

pattern specification, and tissue and organ development and morphogenesis. In contrast,

downregulated GO processes included cell organization and biogenesis, RNA and DNA

metabolism, cell cycle and cytoskeletal organization, and microtubule biogenesis. Compared

to undifferentiated cells that undergo rapid replication and express a wide variety of genes, the

downregulated processes that occur with electrical stimulation are likely due to decreased cell

cycling as the cells acquire a more limited number of genes that are appropriate for the

differentiated state. Thus, RNA processing is reduced, along with cell division and mitotic

processes. Ingenuity Pathway Analysis also demonstrates significant changes in

developmental processes after electrical stimulation (Fig. 8). Based on Ingenuity‟s database of

gene networks, the functional categories that are most significantly changed by electrical

stimulation include gene expression, organismal and organ development, cellular

development, cell cycle, and specific physiological systems such as nervous, hematological,

musculoskeletal, and cardiovascular development.

29

Table 3: Upregulated Gene Ontology (GO) biological processes in electrically-

stimulated mESCs vs. controls

Category Genes in

Category

% of

Genes in

Category

Genes in

List in

Category

% of

Genes in

List in

Category

P-

Value

UPREGULATED

GO:7275: development 2213 13.96 51 22.77 0.000231

GO:9653: morphogenesis 913 5.761 26 11.61 0.000526

GO:48513: organ development 1056 6.663 26 11.61 0.00407

GO:9887: organ morphogenesis 513 3.237 15 6.696 0.00638

GO:9790: embryonic development 219 1.382 9 4.018 0.00402

GO:1501: skeletal development 118 0.745 9 4.018 4.52E-05

GO:7389: pattern specification 182 1.148 8 3.571 0.0044

GO:9888: tissue development 161 1.016 7 3.125 0.008

GO:48598: embryonic

morphogenesis 82 0.517 6 2.679 0.00107

GO:31214: biomineral formation 68 0.429 6 2.679 0.000393

GO:1503: ossification 68 0.429 6 2.679 0.000393

GO:46849: bone remodeling 76 0.48 6 2.679 0.000715

GO:48514: blood vessel

morphogenesis 127 0.801 5 2.232 0.0343

GO:1525: angiogenesis 109 0.688 5 2.232 0.0193

GO:35107: appendage

morphogenesis 38 0.24 5 2.232 0.000185

GO:35108: limb morphogenesis 38 0.24 5 2.232 0.000185

GO:30326: embryonic limb

morphogenesis 37 0.233 5 2.232 0.000163

GO:35113: embryonic appendage

morphogenesis 37 0.233 5 2.232 0.000163

GO:48736: appendage development 38 0.24 5 2.232 0.000185

GO:30900: forebrain development 33 0.208 3 1.339 0.0111

GO:9954: proximal/distal pattern

formation 7 0.0442 2 0.893 0.00399

GO:51216: cartilage development 15 0.0946 2 0.893 0.0185

GO:30111: regulation of Wnt

receptor signaling pathway 10 0.0631 2 0.893 0.00831

GO:30178: negative regulation of

Wnt receptor signaling pathway 7 0.0442 2 0.893 0.00399

GO:35026: leading edge cell

differentiation 1 0.00631 1 0.446 0.0141

GO:48384: retinoic acid receptor

signaling pathway 1 0.00631 1 0.446 0.0141

Genes associated with pluripotency are downregulated with electrical

stimulation. mESC self-renewal is dependent on a core set of transcription factors involved in

the development of the embryo: OCT4, NANOG, SOX2, and FOXD364, 65

. These embryonic

transcriptional factors are essential for the formation and maintenance of the inner cell mass

30

Table 4. Downregulated GO biological processes in electrically-stimulated mESCs vs.

controls

Category Genes in

Category

% of

Genes in

Category

Genes in

List in

Category

% of

Genes in

List in

Category

P-Value

DOWNREGULATED

GO:16043: cell organization and biogenesis 2013 12.7 67 24.1 1.15E-07

GO:16070: RNA metabolism 605 3.818 29 10.43 9.56E-07

GO:6259: DNA metabolism 653 4.12 28 10.07 1.23E-05

GO:7049: cell cycle 784 4.947 27 9.712 0.00065

GO:6396: RNA processing 518 3.269 26 9.353 1.55E-06

GO:6412: protein biosynthesis 723 4.562 21 7.554 0.0166

GO:6397: mRNA processing 421 2.656 21 7.554 1.78E-05

GO:16071: mRNA metabolism 436 2.751 21 7.554 2.99E-05

GO:7010: cytoskeleton organization and biogenesis 526 3.319 20 7.194 0.00102

GO:398: nuclear mRNA splicing, via spliceosome 339 2.139 19 6.835 8.97E-06

GO:8380: RNA splicing 344 2.171 19 6.835 1.10E-05

GO:74: regulation of progression through cell cycle 429 2.707 14 5.036 0.0195

GO:7017: microtubule-based process 204 1.287 11 3.957 0.000973

GO:45786: negative regulation of progression

through cell cycle 110 0.694 7 2.518 0.00325

GO:226: microtubule cytoskeleton organization

and biogenesis 67 0.423 6 2.158 0.00112

GO:79: regulation of cyclin dependent protein

kinase activity 13 0.082 3 1.079 0.00134

GO:7026: negative regulation of microtubule

depolymerization 18 0.114 2 0.719 0.039

GO:7019: microtubule depolymerization 20 0.126 2 0.719 0.0473

GO:31114: regulation of microtubule

depolymerization 18 0.114 2 0.719 0.039

GO:31111: negative regulation of microtubule

polymerization or depolymerization 18 0.114 2 0.719 0.039

GO:31110: regulation of microtubule

polymerization or depolymerization 18 0.114 2 0.719 0.039

GO:85: G2 phase of mitotic cell cycle 6 0.0379 2 0.719 0.00439

GO:51319: G2 phase 6 0.0379 2 0.719 0.00439

during mouse preimplantation development, and for self-renewal of pluripotent ESCs.

Interestingly, two of these factors, OCT4 (POU5F1) and FOXD3, were both downregulated

1.4-fold after electrical stimulation, indicating that important changes are occurring in the

cellular pluripotency programs. Other important downregulated embryonic genes include

UTF1 (undifferentiated embryonic cell transcription factor 1, down 1.4-fold) and RIF1 (Rap1

interacting factor 1, down 1.5-fold). UTF1 is specifically expressed in the inner cell mass and

31

primitive ectoderm and is down-regulated at early primitive streak stages66

. A recent study

suggests a role for UTF1 in the proliferation rate and teratoma-forming capacity of ESCs67

,

and it has been shown to be involved in ESC differentiation68

. RIF1 is an ortholog of a yeast

telomeric protein and is upregulated in mouse ESCs and germ cells. In human cells, RIF1

associates with dysfunctional telomeres and has a role in DNA damage response. Notably,

RIF1 is also a target of OCT4 and NANOG in human ESCs, underlining its functional

importance to ESC biology. Lastly, we observed 1.7-fold downregulation of LIN28, an

Figure 8. Ingenuity Pathways Analysis of transcriptome alteration after electrical

stimulation of mESCs. Among many changes in development, many specific physiological

systems such as the nervous, hematological, musculoskeletal, and cardiovascular

development were noted.

32

important embryonic gene in mouse and human69

that was recently used by the James

Thomson lab to reprogram fibroblasts into pluripotent stem cells70

.

Embryonic germ layer development and body axis specification gene programs

are upregulated after electrical stimulation. One of the earliest commitment steps in

embryogenesis is the formation of the primary germ layers, mesoderm, endoderm, and

ectoderm, which eventually become all somatic cell types in the body. Mesoderm and

endoderm are formed during gastrulation, a process that involves the movement of

undifferentiated epiblast cells through the primitive streak71

. The genes and signals involved in

early embryogenesis and body-axis formation are complex and poorly understood, though a

number of important regulatory genes have been identified and can be used as markers of

development. Many of these developmental genes were slightly upregulated after electrical

stimulation. For example, HOXD10 was upregulated 1.6-fold after electrical stimulation and

is known to regulate a number of embryonic processes, including limb, neuron, and skeletal

morphogenesis, anterior/posterior pattern formation, and nervous system development72, 73

.

Another gene, T-box transcription factor 18 (TBX18, upregulated 1.8-fold), regulates smooth

muscle cell differentiation, somitogenesis (early mesodermal divisions of the vertebrate

embryo), as well as anterior/posterior axis specification74, 75

. Aldehyde dehydrogenase family

1, subfamily A2 (ALDH1A2) and ALDH1A3 were upregulated 1.6- and 1.7-fold,

respectively. The products of these genes catalyze the synthesis of retinoic acid, a signaling

molecule that functions in developing and adult tissues to regulate body axis and cell

differentiation76

. These two genes establish local embryonic retinoic acid levels and therefore

regulate anterior/posterior pattern formation, limb development, and even organ development

such as brain, heart, pancreas, and eye.

Mesoderm and cardiac development. Mesoderm gives rise to cardiac, skeletal, and

smooth muscle, as well as hematopoietic and endothelial cells. We observed a number of early

33

mesodermal genes that were upregulated with electrical stimulation. Significantly, the

mesodermal FOXC1 and FOXC2 genes, upregulated 1.6 and 1.5-fold respectively after

electrical stimulation, are particularly important for heart development and morphogenesis.

FOXC1 and FOXC2 encode closely related transcription factors that are expressed in

mesoderm and endothelial and mesenchymal cells of the developing heart and blood vessels,

including the second heart field, cardiac neural crest cells, and endocardium77, 78

. Another

interesting finding was the upregulation of the transcription factor Islet1 (ISL1, up 1.5-fold),

which has been reported to be a marker of cardiac progenitor cells79, 80

. Though there is not yet

total agreement in the field, it is generally accepted that ISL1, which is expressed in the

secondary heart field, directly regulates cardiac precursor cells. Further, ISL1 has been shown

to be expressed by progenitors of the outflow tract, right ventricle, and a majority of atrial

progenitors79

. It should be noted that ISL1 is not cardiac-specific since it is also expressed in

motor neurons and pancreas during embryogenesis, as well as in normal adult islet cells81

, so

its initial activation and its actions on downstream targets likely require combinatorial

mechanisms and influences82

.

A number of studies have shown that stage-specific inhibition of canonical WNT

signaling is required for cardiac development, so the upregulation of the WNT signaling

pathway antagonist Dickkopf homolog 1 (DKK1, up 2.0-fold) is of particular interest. DKK1

is well known to specify cardiac mesoderm when expressed at the appropriate stage of

development, and application of this factor has recently been shown to encourage in vitro

differentiation of human ESCs into cardiac progenitors83

. Other cardiac-related genes that

were upregulated after electrical stimulation include the ATPase, Ca++ transporting, cardiac

muscle, fast twitch 1 (ATP2A1, upregulated 1.4-fold), and SEMA6D (upregulated 1.7-fold),

which regulates endocardial cell migration in the developing heart and spinal cord84, 85

.

34

Musculoskeletal Development. Closely related to cardiac muscle are the

musculoskeletal tissues: both systems are derived from mesoderm and share many common

genes and pathways during development. Several significant musculoskeletal-related GO

biological processes were upregulated after electrical stimulation, namely skeletal

development (GO 1501), ossification (GO 1503), bone remodeling (GO 46849), cartilage

development (GO 51216), as well as numerous limb and appendage processes. Specific genes

of interest include the matrix metalloproteinase MMP-13 (upregulated 2.4-fold), which is

involved in the formation and remodeling of bone86

and is expressed in chondrocytes,

osteoblasts, and periosteal cells in early development87

. Another gene, SRY-box containing

gene 9 (SOX9, upregulated 1.6-fold), is a transcription factor that is expressed in chondrocytes

(in addition to central nervous and urogenital systems), and is a regulator of the chondrocyte

differentiation and cartilage formation88

. Runt related transcription factor 2 (RUNX2,

upregulated 1.5-fold) is a regulator of skeletal development89

, and is a potent inducer of late

stages of chondrocyte and osteoblast differentiation90

. The MYOD gene has a central role in

regulating muscle differentiation, and we observed 2-fold downregulation of an inhibitor of

this gene, the transcription factor MYOD family inhibitor (MDFI)91

, though we did not

observe concurrent upregulation of MYOD itself. Lastly, paired-like homeodomain

transcription factor 1 PITX1 (upregulated 1.7-fold), has been shown to be important for

skeletal and cartilage development, as well as hindlimb morphogenesis92

.

Nervous System Development. We also observed upregulation of genes and

pathways involved in nervous system development after electrical stimulation. Perhaps the

most striking example is the upregulation of neural cell adhesion molecule 1 (NCAM1,

upregulated 1.5-fold), a common marker of primitive neuroectoderm. Another gene,

Delta/NOTCH-like EGF-related receptor (DNER, upregulated up 1.7-fold), is a neuron-

specific transmembrane protein that functions as a ligand of NOTCH during cellular

35

morphogenesis of glia in the mouse cerebellum93

. CCAAT/enhancer binding protein beta

(CEBPB, upregulated 1.5-fold) has been implicated in neuron differentiation94

as well as

adipogenesis95

. Twisted gastrulation gene 1 (TWSG1, upregulated 1.4-fold) plays a number of

regulatory roles in embryonic development, including forebrain development and mesoderm

via BMP signaling96

. Lastly, FOXD1 (upregulated 1.8-fold) plays known roles in axon

guidance97

as well as kidney development98

.

Taken together, we have highlighted a number of genes and processes that, though the

overall fold changes are not dramatic, seem to indicate a trend towards development and

differentiation after electrical stimulation of murine EBs. Simultaneous to this mild

upregulation of mature gene programs is the apparent downregulation of important self-

renewal and pluripotency genes, including OCT4 and FOXD3. Some of the most significant

physiological systems that exhibited altered gene expression were the cardiovascular,

neurological and musculoskeletal systems. Since the cells in these particular tissues and

organs are electrically excitable, it is tempting to attribute these changes to the external

electrical stimulation of EBs. Studies such as this give new insights into directed

differentiation of embryonic stem cells, and provide evidence that artificially mimicking the in

vivo environment of a differentiating stem cell during development may influence its lineage

preference in an in vitro system as well.

36

2. DIFFERENTIATING THE PLURIPOTENT STATE

Transcriptional profiling of human embryonic stem cell-derived endothelial cells23

.

INTRODUCTION

Ischemic heart disease is a leading cause of mortality and morbidity worldwide.

Current treatments fail to address the underlying scarring and cell loss, which is a major cause

of heart failure after infarction99

. Therapeutic angiogenesis and/or vasculogenesis using

cellular transplantation is a promising and novel strategy to increase blood flow in patients

with severe ischemic heart disease and peripheral vascular disease100

. With their capacity for

unlimited self-renewal and pluripotency, human embryonic stem cells (hESCs) may provide

an alternate source of therapeutic cells by allowing the derivation of large numbers of

endothelial cells for various tissue repair and cell replacement therapies.

We designed a protocol that involved EB formation and expansion of the endothelial

lineage by sub culturing EBs in collagen. We then derive a highly pure endothelial population

that is CD31/CD144+ (two common endothelial surface markers101

) after double sorting using

flow cytometry. In order to define at a molecular level the changes occurring at each stage of

hESC differentiation to endothelial cell progeny, and to validate that these cells are similar to

human umbilical vein endothelial cells (HUVECs), we performed transcriptional profiling

using whole human genome microarrays. Taken together, we report that hESC-ECs express

appropriate patterns of endothelial genes, form functional vessels in vivo, and improve cardiac

function. These results suggest that hESC-ECs may provide a novel therapy for ischemic heart

disease.

37

MATERIALS AND METHODS

Microarray hybridization and data acquisition. Total RNA was isolated from

undifferentiated H9 hESCs, day 12 EBs, hESC-ECs (after CD31/CD144 sort), and HUVECs.

Four samples from each group (for a total of 16 unique samples) were harvested for RNA

isolation. Using Agilent Low RNA Input Fluorescent Linear Amplification Kits, cDNA was

reverse transcribed from each of 16 RNA samples representing four biological quadruplicates,

as well as the pooled reference control, and cRNA was then transcribed and fluorescently

labeled with Cy5/Cy3. cRNA was purified using an RNeasy kit (Qiagen, Valencia, CA, USA).

825 ng of Cy3- and Cy5- labeled and amplified cRNA was hybridized to Agilent 4×44K

whole human genome microarrays (G4112F) and processed according to the manufacturer‟s

instructions. The image files were extracted using Agilent Feature Extraction software version

9.5.1 applying LOWESS background subtraction and dye-normalization. The data were

analyzed using GeneSpring GX 7.3.1. (Agilent Technologies, Santa Clara, CA) with multiple

testing correction to identify genes which had statistically significant changes in expression

between stages, and K-means clustering to identify clusters of genes having unique temporal

expression profiles. For hierarchical clustering, we used Pearson correlation for similarity

measure and average linkage clustering. Gene Ontology (GO) overrepresentation analysis was

performed also using the GeneSpring software.

RESULTS

Major changes in gene expression between stages highlight developmental

progression of the endothelial lineage. A summary of our major findings is shown in Fig.

9A. To obtain an overview of the transcriptional landscape, we looked at the data using

principal components analysis (PCA), a dimensional reduction technique which identifies

"principal components" or major trends in gene expression in the overall data (Fig. 9B). PCA

38

39

Figure 9. Transcriptional profiling of hESC-derived endothelial cells. (A) hESCs express

high levels of pluripotency-associated genes including OCT4, SOX2, NANOG, Lefty, and

DNMT. At the EB stage, the cells express high levels of mesodermal master regulators such

as Tbx2, and BMP4 as well as highly enriched levels of endothelial specific master

regulators including EPAS1, TIE2, KDR, and EFNA. This population also expresses genes

from other cell layers, and many developmental genes from WNT and homeobox families.

hESC-ECs downregulate early mesodermal genes and express more endothelial specific

genes, while HUVECs have the highest levels of mature endothelial gene expression with

very few other developmental lineages represented. (B) Principal Components Analysis

(PCA) shows that replicate experiments of each cell type are very similar while

differentiation groups separate significantly along components 1 and 2. (C) Hierarchical

Clustering Analysis - Cells from each developmental stage cluster relatively close to each

other, with the most distance between hESCs and HUVECs. (D) K-means clustering analysis

identifies major trends in gene expression across the time course.

demonstrates that each of the four replicates from each stage has very similar transcriptional

profiles to one another, but distinctly different between stages, as expected. Hierarchical

clustering of the microarray experiments (Fig. 9C) likewise shows that the overall gene

expressions among replicates of each stage are very similar, with progressive differences

between more distantly separated stages. Of particular interest is the close clustering of hESC-

ECs and HUVECs, indicating a high degree of similarity between their respective expression

profiles.

K-means clustering reveals broad trends in gene expression. Given the enormous

amount of data in microarrays, it is often useful to first obtain an overview of the broader

trends in the data before proceeding to specific genes. Using K-means clustering analysis, we

can easily identify four major trends in the microarray data (Figure 9D). Cluster 1 is

composed of 5,089 genes whose expression is increased at the hESC-EC and HUVEC stages

compared to hESC and EB (Figure 9D). To better understand which cellular processes are

important in this cluster, we performed statistical Gene Ontology (GO) biological process

overrepresentation analysis. GO analysis of cluster 1 shows that the processes to which these

genes contribute include many basic differentiated cell functions such as protein and

40

macromolecule metabolism, cell cycle, vesicle-mediated transport, cytoskeletal biogenesis,

and control of apoptosis. The converse expression pattern is seen in the 3,998 genes

composing cluster 2, which are downregulated in the hESC-CM and HUVEC stages compared

to hESCs and EB. The processes overrepresented in this cluster primarily involve nucleic acid

synthesis, DNA replication and chromatin maintenance, cell cycle and mitosis, and

transcription in general. This theme reflects the fact that earlier undifferentiated cells are

undergoing rapid replication and production of broad ranges of transcripts, while cell cycling

slows dramatically later in development as cells begin to express a more limited number of

genes that are appropriate for the differentiated state.

Cluster 3 is comprised of 2,965 genes whose expression increases when hESCs

transition to the EB stage, then progressively decrease in the hESC-EC and HUVEC stages.

The overrepresented processes in this cluster correspond to development and differentiation

pathways – the critical steps for EBs, including many aspects of development: organ, skeletal,

nervous system, and morphogenesis. The final interesting cluster of genes is cluster 4, which

represents 2,952 genes whose expression progressively increases from the hESC to EB to

hESC-EC to HUVEC stages. Similarly to cluster 1, many of the overrepresented biological

processes are related to basic differentiated cell pathways such as protein transport and cellular

localization. However, the endothelial cells processes (blood vessel morphogenesis and

development, angiogenesis and vasculature development) are also highly represented here.

Cluster 4 thus describes those specific processes important for endothelial cell development

and differentiation; we will analyze these specific pathways and genes in more detail below.

hESCs exhibit unique biologic processes and molecular signature. GO analysis

reveals that the most highly upregulated processes in hESCs are almost exclusively cell

cycling and mitosis, as well as nucleic acid synthesis and metabolism. This was not surprising

given that hESCs‟ primary mission is to self renew. hESCs are also characterized by a

41

network of genes important for pluripotency, including the unique homeobox transcription

factor NANOG, which is the main downstream effector of this network. When we compared

expression patterns in hESCs to EB cells, we found that there were 2,170 genes expressed

much more highly in hESCs. The most dramatically elevated transcript in hESCs was

NANOG, which is expressed at a level 60 times higher in hESCs than in EB102, 103

. POU5F1

(also known as OCT4), upstream of NANOG, is one of the critical regulators of pluripotency

in the mammalian embryo, and our results show that it is expressed 37-50 fold more highly in

the undifferentiated hESCs when compared to EB. SOX2, another key pluripotency gene, is

expressed at 14.5 times the level in hESCs as in EB104

. Other known markers of ESC status

are also clearly present at high levels: TDGF1 and 3 (CRYPTO1 and 3)105

, expressed >10 fold

higher levels in hESCs; LIN28 (7 fold higher); the ESC markers such as developmental

pluripotency-associated 4 (DPPA4) (8.5 fold)106

and homeobox expressed in ESCs 1 (HESX1,

8 fold)107

, and left-right determination factor 1 (LEFTY1, up 47 fold), galanin (GAL, up 23

fold), DNA (cytosine-5-)-methyltransferase 3 beta (DNMT3B, up 19 fold), fibroblast growth

factor 2 (basic) (FGF2, up 7 fold).

EB cells express many mesodermal and endothelial specific gene programs.

Differentiation to the EB stage is a critical step as development occurs spontaneously down

many paths. Analysis of the biological processes upregulated in EBs reflects this

heterogeneity, with overrepresentation of vasculature and blood vessel development, skeletal

development, lung development, and others in the microarray data. We noted many

angiogenic and endothelial-specific genes that were upregulated in EBs, including the

endothelial transcription factor endothelial PAS domain protein 1 (EPAS1, also called

hypoxia-inducible factor-2, up 2.1 fold)108

, endothelial TEK tyrosine kinase (TEK, also known

as TIE2, up 12 fold)109

, kinase insert domain receptor (KDR, up 5-8 fold), the ephrin receptor

protein-tyrosine kinases EFNA1 (up 5-14 fold) EFNB2 (up 3 fold)110-112

, endothelial protein C

42

receptor (EPCR , up 2-3 fold)113

, and the claudin family member CLDN5 (up 4.2 fold) that is a

component of endothelial and epithelial tight junctions114

. VEGF variant 1 (VEGF, up 2.4

fold) and VEGFC (up 11.8 fold) were also upregulated in EBs. Interestingly we also observed

upregulation in EBs of the estrogen receptor GPR30 (up 20 fold), as well as a number of blood

coagulation genes indicating hematopoiesis, including plasminogen (PLG, up >5 fold), tissue

plasminogen activator (PLAT, up 30 fold), and coagulation factors II (F2, also known as

thrombin, up 56 fold), III (F3), VII (F7, up 8.5 fold), and X (F10, up 5 fold), among others.

The coagulation cascade has been shown to modulate endothelial cell function in developing

blood vessels, and thrombin's action on endothelial cells contributes to vascular development

and hemostasis115

. Lastly, confirming previous reports that the WNT signaling pathway is

important for stem cells and endothelial commitement116, 117

, we observed upregulation of

WNT5A (up 9.2 fold) and WNT5B (up 5 fold).

We do still see at this stage expression of genes characteristic of all three cell layers,

endoderm, mesoderm and ectoderm. These include endodermal genes such as FOXA1

(HNF3A, up 11 fold)118, 119

, the mesodermal genes such as the musculoskeletal transcription

factors runt-related transcription factor 2 (RUNX2, up 8 fold)120

and SOX9 (up 5.1 fold),

cardiogenic/mesodermal and endodermal zinc finger transcription factors GATA4 (up 2.8

fold), GATA5 (up 17.4 fold) and GATA6 (up 12.6 fold)121-123

, and heart and neural crest

derivatives expressed 1 (HAND1, up 9.6 fold), as well as the neuro-developmental gene

CLN5 (up 3.3 fold)124

, demonstrating that EBs still contain cells from each lineage. Several

hemoglobin genes (HBG1, HBE1, HBZ) are also upregulated in EBs, indicating development

of erythropoietic cells. We also noted upregulation of the multi-functional factors bone

morphogenetic proteins 4 (BMP4, up 6.3 fold) and 5 (BMP5, up 14.8 fold), LIM domain only

2 (rhombotin-like 1) (LMO2, up 7.2 fold), the transcription factor forkhead box P1, transcript

variant 1 (FOXP1, up 2.3 fold). Many developmental regulatory genes such as the homeobox

43

genes (HOXA3, A4, A9, A10, B2, B3, B4, B5, B8, B9, B13, and D1)125

and limb bud and

cardiac T-box family of transcription factors TBX2 and TBX3126, 127

were also upregulated,

further illustrating the multiplicity of developmental lineages in EBs.

hESC-ECs downregulate early developmental genes and upregulate endothelial

genes. Differentiation of EBs to endothelial cells is marked by a downregulation of processes

such as transcription, non-endothelial developmental pathways (including lung, skeletal and

nervous system), and blood coagulation. In looking at the GO pathways that are

overrepresented in hESC-ECs, apoptotic pathways are significantly upregulated, as well as

angiogenesis, blood vessel and vasculature development. The downregulation of

developmental programs is evidenced by the decreased expression of a number of homeobox

genes (HOXB2, 3, 4, HOXA2, HOXD1, IRX5, HESX1), as well as the limb bud

developmental regulators PITX1 and PITX2128

. Also noteworthy are the mesodermal

transcription factors GATA5, GATA6, MEF2C and FOXC1, whose expression decreases

between the EB and EC stages indicating that hESC-ECs have progressed beyond the

mesodermal stage. Nevertheless, there is still transcriptional evidence for the presence of other

mesodermal derivatives such as hematopoietic derivatives with the increased expression of

skeletogenic genes such as RUNX2 (6.6 fold).

Interestingly, we also observed downregulation of many transforming growth factor-

beta (TGF-beta) superfamily signaling proteins, including Noggin (NOG, down 3 fold),

TGFB3, and BMP2, BMP4, and BMP5, suggesting that this pathway is less important for

endothelial development as these cells mature. Interestingly, the endothelial-specific receptor

NOTCH4 (down 2.3 fold) and its ligands delta-like 1 (DLL1) and jagged 2 (JAG2) were all

downregulated, indicating a reduction in NOTCH signaling that is required for early cell fate

determination129, 130

. However there is still evidence of continued developmental programs that

are active in hESC-ECs compared to EBs, as demonstrated by the upregulation of BMP1,

44

BMP6, and the homeobox genes HOXA7, A9, A10, and C9. Interestingly, the chemokine (C-

X-C motif) receptor 4 (CXCR4, down 21 fold), which is important for chemotaxis of

endothelial cells, is down in hESC-ECs compared to EBs.

Concurrent with the downregulation of early developmental gene programs, we

observe substantial increases in expression of many vasculature-related genes in hESC-ECs

compared to EBs. These genes include the endothelial-specific transcription factors Kruppel-

like factor 2 (KLF2, up 3.5-5 fold)131

and EPAS1 (up 2.3 fold), PECAM1 (up 3.1 fold),

endothelial lipase (LIPG, up 2.9-3.8 fold), and endothelin 1 (EDN1, up 16 fold). A number of

genes demonstrated continued upregulation from the hESC to EB to EC stages, including

endothelial protein C receptor (EPCR, up 2.3 fold) peroxisome proliferative activated

receptor, gamma (PPARG, up 12.2 fold), and endothelial cell adhesion molecule (ESAM, up

2.5-3 fold). In the endothelium, caveolin-1 (CAV1, up 23-31 fold) regulates nitric oxide

signaling by binding to and inhibiting endothelial nitric oxide synthase (eNOS)132

. GATA2

expression was increased 4.6 fold and is known to be involved with regulation of PECAM1,

endothelin-1, and ICAM2133

. The endothelial-specific gene ROBO4 (up 2.6-8.1 fold)

maintains vascular integrity and inhibits endothelial cell migration and angiogenesis134

. Other

genes upregulated in hESC-ECs include the tyrosine kinase receptor for both VEGF and

semaphorin, neuropilin 1 (NRP1, up 3-5 fold), that is expressed in both endothelial and

neuronal cells135

, apelin (APLN, up 2-20 fold), ecto-5'-nucleotidase (CD73, also known as

NT5E, up 20-26 fold)136

, ITGB3 (up 7-12 fold), and CD109 antigen (up 3-4 fold). Lastly,

compared to EBs, hESC-ECs further upregulate of WNT5A and WNT5B (up 2.4 and 3.6 fold,

respectively), but downregulate of WNT2B, 3, 6, and 11. Of note, the WNT signaling pathway

inhibitor dickkopf homolog 1 (DKK1, up 26 fold) is upregulated in hESC-ECs, which has

recently been shown to be important for in vitro differentiation of hESCs to cardiovascular

progenitors83

.

45

Significant changes between hESC-ECs and HUVECs. One of our goals in this

study is to compare our EC population to a cell population that would likely be optimal for

cell transplantation into ischemic tissue. An optimal cell type would be committed to the

endothelial lineage but would still retain the capacity to undergo mitosis and thereby

regenerate damaged vasculature. Assessing the similarities, as well as differences, in gene

expression between hESC-ECs and HUVECs will thus give insight into the overall maturation

of our differentiated cells. Many of the increased processes in HUVECs include angiogenesis,

blood vessel morphogenesis and development, iron ion sequestering and transport, and

phosphorylation, suggesting that HUVECs may be a somewhat more differentiated or

“matured” endothelial cell. We were curious about the increased iron metabolism, and found

that recent evidence has uncovered a mechanism for feedback regulation of EPAS1 translation

in response to iron, suggesting a relationship between iron levels and EPAS1 regulation137

.

Downregulated processes in HUVECs reflect the continued differentiation of HUVECs when

compared to hESC-ECs, including skeletal and nervous system development, as well as a

reduction in VEGF and WNT receptor signaling. Specific genes that are downregulated in

HUVECs include RUNX2 (down 320 fold) and SOX9 (down 49 fold), as well as the

hemoglobin genes (HBG1, HBE1, HBZ), indicating that there is further differentiation

towards endothelial rather than musculoskeletal of erythropoietic lineages. Compared to

hESC-ECs, HUVECs markedly downregulate WNT5A (down 256 fold) and WNT5B (down

91 fold), and to a lesser degree WNT3A (down 2.9 fold), WNT3 (down 2.1 fold), WNT6

(down 3 fold), and WNT11 (down 3.2 fold). As in EBs, the estrogen receptor (GPR30, up 7.5

fold) is again upregulated in HUVECs, and may contribute to mobilization of endothelial

cells138

.

A number of early developmental genes are further downregulated in HUVECs. These

include the mesodermal transcription factors NKX2.5 (down 5.6 fold) and HAND2 (down 6.7

46

fold). Bone morphogenetic protein regulation is alternately regulated depending on the BMP.

For example, HUVECs downregulate BMPR1A 15-20 fold and BMP5 3 fold, yet upregulate

BMP4 (up 7.8 fold) and BMP6 (up 14.4 fold). Likewise, homeobox genes do not exhibit a

clear pattern of regulation in HUVECs vs. hESC-ECs, where HOXA2, A3, A4, A5, A7, B4,

D1, and D3 are upregulated while HOXB6, B9, B13, and C9 are downregulated. The early

developmental T-box transcription factors, TBX1, TBX2, TBX3, are also all downregulated in

HUVECs. A number of forkhead box transcription factors were downregulated (Fox A1, A2,

A3, D1, F1, F2), with FOXD1 (down 38 fold) and FOXA1 (down 19-34 fold) the most

dramatically downregulated. Interestingly, FOXC1 was the only forkhead transcription factor

that was upregulated (6.6 fold) in HUVECs.

In contrast, HUVECs further upregulate a number of important genes related to

angiogenesis and vasculature development. These genes include the intercellular adhesion

molecule KDR (up 6.2-7.5 fold), fms-related tyrosine kinase 4 (FLT4, up 5.2 fold), ICAM2

(up 51 fold), CLDN5 (up 8.7-10.3 fold), ROBO4 (up 6-7 fold), von Willebrand factor (VWF,

up 97 fold), EPCR (up 4 fold), PECAM1 (up 6.9 fold), nitric oxide synthase 3 (NOS3, up 6.7

fold), EFNA1 (up 5-10 fold), ITGB3 (up 4-5 fold), APL (up 3-8 fold), EPHA4 (up 5-14

fold)139

, cadherin 5, type 2, VE-cadherin (CDH5, up 6.5 fold), the tight junction protein 2

(TJP2, up 4.5 fold), endothelial cell-specific molecule 1 (ESM1, up 7.5 fold), angiopoietin 2

(ANGPT2, up 2.9 fold – this is in contrast to angiopoietin 1, ANGPT1, which is

downregulated 13.5 fold), endothelin 1 (EDN1, up 6.4 fold), TEK (or TIE2, up 4.5 fold),

NOTCH1 (up 4 fold), CD109 antigen (Gov platelet alloantigens) (CD109, up 5.4 to 6.4 fold).

The chemokine (C-X-C motif) receptor 4 (CXCR4, up 24 fold) is important for chemotaxis of

endothelial cells. Taken together, HUVECs likely represent a slightly more differentiated

endothelial cell type compared to hESC-ECs.

47

CONCLUSION

Here we present a differentiation procedure that promotes differentiation of hESCs

into endothelial lineage cells. Global gene expression profiling of hESCs during directed

endothelial differentiation revealed significant enrichment of a number of endothelial genes

and appropriate biological processes that correspond to each differentiation stage (note:

expression of selected genes were confirmed using qPCR analysis, see Li et al.23

for data).

Though hESC-ECs and HUVECs revealed a high degree of similarity between their respective

expression profiles, hESC-ECs do not resemble HUVECs completely. Previous reports have

shown that hemodynamic forces such as fluid shear stress can affect gene transcription in vivo

and in vitro140, 141

. This suggests that shear stress on HUVECs may contribute to the

discrepancy of expression profiles between the mature HUVEC and the more progenitor-like

hESC-EC, which has not been exposed to hemodynamic forces during in vitro differentiation.

Thus, to generate more functional and mature endothelial cells, shear stress conditioning of

hESC-ECs may be needed in the future. Importantly, our microarray analysis provides a

greater understanding of the patterns of activation of specific genes during endothelial

differentiation, and identifies novel gene targets that may be used to enhance endothelial

differentiation in the future.

48

2. DIFFERENTIATING THE PLURIPOTENT STATE

The miR-145 pluripotency loop in human embryonic stem cells142

.

The discovery and isolation of human embryonic stem cells (hESCs) has given

medical science the tantalizing prospect of one day regenerating organs and tissues in human

patients. In contrast to multipotent adult stem cells, hESCs have the advantage of being

pluripotent, which endows them with the ability to differentiate into virtually every cell type in

the human body. However, forcing hESCs to change their phenotype is an imperfect science,

and is often time-consuming, resource-intensive, and plagued by poor yields. Fundamentally,

what is needed is better control over the molecular programs that induce, maintain, and repress

differentiation. It has been well established that pluripotency and self-renewal, the twin pillars

of embryonic stem cell biology, are regulated by a core network of transcription factors,

including OCT4, SOX2, NANOG, and KLF4, which are integrated into complex molecular

circuits. In fact, over-expression of these and other core factors can induce pluripotency in

human somatic cells, as was demonstrated by Shinya Yamanaka and James Thomson in 2007.

Mirroring the difficulties with differentiating hESCs, however, reprogramming efficiencies

have remained static at ~0.01%, highlighting the fact that scientists still lack great control over

cell fate switching. The question now is, what‟s missing? What other elements should be

considered in order to achieve efficient differentiation and de-differentiation of cells?

In an important recent study by Xu et al.143

, one potential regulatory mechanism of

hESCs was found to be microRNAs (miRNAs), and specifically miR-145. These small,

noncoding RNAs play important posttranscriptional regulatory roles by targeting messenger

RNAs (mRNAs) for cleavage or translational repression144

(Fig. 10), and are key components

of an evolutionarily conserved system of RNA-based gene regulation in eukaryotes145

. In the

past five years, a significant body of evidence has identified miRNAs as major arbiters of

49

Figure 10. MiRNA biogenesis and targeting. Primary transcripts of miRNAs are transcribed

as individual miRNA genes, from introns of protein-coding genes, or from polycistronic

transcripts. These are cleaved into 70–100 nucleotide, hairpin-shaped precursors, exported to

the cytoplasm, and further cleaved into an miRNA:miRNA* duplex. Assembled into the

RNA-inducible silencing complex (RISC), the mature miRNA negatively regulates gene

expression by either translational repression or mRNA degradation.

50

hESC biology, and specifically of fate switching. hESCs are known to express miRNAs that

are often undetectable in adult organs146

, and Dicer- and DGCR8-deficient murine ESCs,

which have defects in miRNA generation, are characterized by reduced proliferation and

failure to differentiate147, 148,149

. A recent study has attempted to incorporate miRNA gene

regulation into a model of transcriptional regulatory circuitry of murine ESCs by generating

genome-wide maps of binding sites for key ESC transcription factors such as OCT4, SOX2,

and NANOG150

. These ESC transcription factors were found to bind at many start sites of

miRNA transcripts that have been detected in ESCs, such as the miR-302 cluster. Clearly,

these studies and others have raised the profile of miRNAs in the context of ESC pluripotency,

but have also raised a number of questions. Foremost among these questions is how to identify

specific miRNAs that are directly involved in ESC pluripotency. Of the ~650 human miRNAs

that have been described, only a fraction has actually been annotated; the great majority of

miRNAs have poorly-defined roles in cell fate decisions.

Building on this prior evidence, and to address some of the questions it has raised, Xu

et al. showed through a rigorous series of experiments that one miRNA in particular, miR-

145, directly targets the core ESC factors OCT4, SOX2, and KLF4. Their study began with

the initial observation that miR-145 was significantly upregulated during differentiation of

hESCs into embryoid bodies. After determining that miR-145‟s predicted targets include

OCT4, SOX2, and KLF4, they then used luciferase assays to show that endogenous miR-145

can repress the 3‟ UTRs of the three factors. Furthermore, over-expression and inhibition of

miR-145 in cells resulted in reciprocal expression patterns of the same core factors. To

determine functional effects of miR-145, the authors employed a number of assays, including

flow cytometry for SSEA4 (pluripotency) and propidium iodide (cell cycle) staining, as well

as colony formation, and discovered that miR-145 over-expression in hESCs inhibits self-

renewal and induces differentiation. In contrast, inhibition of miR-145 reduced hESC‟s

51

capacity for differentiation. Interestingly, they also found an OCT4 binding motif 1 kb

upstream of the transcription start site of miR-145, and functionally validated this OCT4

binding site with luciferase reporters, electrophoretic mobility shift assays, and chromatin

immunoprecipitation. Xu et al. conclude that miR-145 regulation of the core pluripotency

factors is a “double-negative feedback loop”: miR-145 targets OCT4, SOX2, and KLF4, but is

itself negatively regulated by OCT4.

With the evidence presented in this exciting study, can manipulation of miR-145 now

be used as a basic tool for differentiating hESCs? Conversely, could inhibition of its function

augment pluripotency induction in somatic cells? Recent work by Cordes et al. (2009)151

has

shown that miR-145, along with miR-143 with which it was found to be co-transcribed, can

promote differentiation and repress proliferation of smooth muscle cells. These two separate

studies therefore each implicate miR-145 as an important control point for cell fate switching.

However, one should remain wary of any claims that miR-145 is the central linchpin for

inducing differentiation. This caution is justified in large measure by the lessons from systems

biology, in which understanding biology at the systemic level, rather than its individual parts,

is the central dogma. We now commonly hear words such as “circuits”, “networks”, and

“programs” to describe molecular regulation of cellular processes. Within these larger

networks are smaller, more focused “loops” such as “incoherent feed forward loops”

(miRNAs and their target genes that have positively-correlated expression) and “double-

negative feedback loops” (as reported by Xu et al. and discussed earlier). While these terms

bring a language of engineering sophistication to biological research, what they imply is that

no single miRNA, mRNA, or any other molecule is solely responsible for a given biological

process. The major role of miRNAs in ESCs may instead be to modulate gene expression

through gene networks in which they are co-expressed with their target genes, thus acting as

genetic buffers of key pluripotency proteins152

. Acknowledging this complex and at times

52

seemingly contradictory relationship between miRNAs and mRNAs, Xu et al. reason that the

coexistence of activation and repression molecules in balance might provide a layer of control

over the dosage of pluripotency factors that maintain the self-renewal state.

The limited annotation of miRNAs, combined with a general lack of knowledge of

their transcriptional start sites, promoter regions, and downstream mRNA targets, have made it

difficult to comprehensively integrate miRNAs into a broader understanding of molecular

networks. This new work by Xu et al. is thus critical for filling in the significant holes in our

understanding of genes and miRNAs in hESCs. As a final example of the complexities

involved in untangling miRNA/mRNA interactions, Xu et al. used TargetScan (528 predicted

targets for miR-145) and Miranda (977 predicted targets) for their initial target prediction

analysis of miR-145. The authors went on to validate just three of these targets, OCT4, SOX2

and KLF4. This begs the question: what about the >500 others? How do these hundreds of

other potential targets affect the intricate orchestration of miRNA/mRNA interactions in

hESCs, and what role, if any, might they also play in self-renewal and pluripotency? Welcome

to the dawn of the miRNA era in embryonic stem cell biology.

53

2. DIFFERENTIATING THE PLURIPOTENT STATE

Dynamic microRNA expression programs during cardiac differentiation of

human embryonic stem cells.

ABSTRACT

MicroRNAs (miRNAs) are a newly discovered endogenous class of small noncoding RNAs

that play important posttranscriptional regulatory roles by targeting messenger RNAs

(mRNAs) for cleavage or translational repression. Human embryonic stem cells (hESCs) are

known to express miRNAs that are often undetectable in adult organs, and a growing body of

evidence has implicated miRNAs as important arbiters of heart development and disease. To

better understand the transition between the human embryonic and cardiac “miRNA-omes”,

we report here the first miRNA profiling study of cardiomyocytes derived from hESCs

(hESC-CMs). Analyzing 711 unique miRNAs, we then identify several interesting miRNAs,

including miR-1, miR-133, and miR-208, that have been previously reported to be involved in

cardiac development and disease and that show surprising patterns of expression across our

samples. We also identify novel miRNAs such as miR-499 that are strongly associated with

cardiac differentiation, and which shares many predicted targets with miR-208. Over-

expression of miR-499 and miR-1 resulted in upregulation of important cardiac myosin heavy

chain genes in embryoid bodies; miR-499 over-expression also caused upregulation of the

cardiac transcription factor MEF2C. Taken together, our data give significant insight into the

regulatory networks that govern hESC differentiation, and highlights the ability of miRNAs to

perturb, and even control, the genes that are involved in cardiac specification of hESCs.

54

INTRODUCTION

Human embryonic stem cell–derived cardiomyocytes (hESC-CMs) hold great promise

for myocardial regeneration after infarction. A number of groups have reported successful

transplantation of hESC-CMs into rodent models of myocardial infarction20, 100, 153

. However, a

significant obstacle still exists with consistent derivation of purified hESC-CM populations. In

order to precisely drive differentiation of hESCs into cardiac cells, a more comprehensive

understanding of the regulatory networks that are involved in this process is needed. hESC-

CM transcriptional profiling has already been reported20

, however no group has yet focused on

the rapidly evolving field of microRNAs (miRNAs) and their expression profiles in hESC-

CMs.

MiRNAs play important posttranscriptional regulatory roles by targeting mRNAs for

cleavage or translational repression144

. Hundreds of human miRNAs have been discovered,

and each is predicted to target tens or hundreds of different mRNAs. hESCs are known to

express miRNAs that are often undetectable in adult organs16, 146, 154

, and some of these

miRNAs are believed to regulate ESC self-renewal and differentiation147

. A growing body of

evidence has also implicated miRNAs in heart disease and development155

. As examples,

miRNAs -21, -195, and -208 have demonstrated roles in heart failure, hypertrophy, and

response to stress156-158

, while miRNAs -1 and -133 are known to regulate cardiac

development, differentiation, and some aspects of disease159-162

. Although previous studies

have highlighted miRNA profiles of hESCs and embryoid bodies16, 154

, to date the miRNA

signature of hESC-CMs remains unknown.

To better understand the transition between the human embryonic and cardiac

“miRNA-omes”, we report here the miRNA profiles of cardiomyocytes derived from human

embryonic stem cells. Our results confirm the presence of a signature group of miRNAs that is

upregulated in hESCs, and whose expression is significantly altered during differentiation to a

55

cardiac lineage. We also identify novel patterns of miRNA expression, as well as individual

miRNAs such as miR-499, that suggest a central role for miRNAs in pluripotency,

differentiation, and maintenance of the cardiac phenotype.

MATERIALS AND METHODS

Culture of undifferentiated hESCs. All studies used cells derived from the female H7 hESC

line (Wicell, Madison, WI) and were maintained in the undifferentiated state using mTeSR™1

medium (StemCell Technologies, Vancouver, Canada).

Preparation of hESC-CMs. Prior to differentiation, hESCs were switched to mouse

embryonic fibroblast conditioned medium (MEF-CM). For these experiments, hESC-CMs

were generated using a method adapted from a previously reported protocol for efficient

cardiogenesis100

.

Immunohistochemical analysis. Undifferentiated hESCs and hESC-CMs were plated onto 8

chamber slides and fixed with acetone on ice for 20 minutes, then stained for

immunofluorescence with the appropriate antibodies. Microscopy was performed using a

ZEISS Axiovert microscropy (Sutter Instrument).

FACS analysis. Cultures were dissociated with 0.05% trypsin, permeabilized with methanol

(and incubated with alpha- sarcomeric actinin (Sigma A7811) or isotype control) in 20% heat

inactivated goat serum with rat anti-mouse Fc block. Samples were washed, incubated with

the appropriate secondary antibody (0.5ug/106 cells), and incubated at room temperature for

15 min in the dark. Samples were washed with 4 ml PBS and resuspended in 200 ul of 0.5%

PFA prior to analysis on a FACSCalibur machine.

56

Lentiviral production and generation of stable microRNA+ hESC lines. Functional

experiments were performed with Lenti-miR™ MicroRNA Precursors and a miRZip anti-

miR-499 microRNA construct (System Biosciences, Mountain View, CA). All experiments

were performed a minimum of three times. SIN lentivirus was prepared by transient

transfection of 293T cells. hESCs were stably transduced with Lenti-miR™ MicroRNA

Precursors or miRZip anti-miR-499 microRNA construct at a multiplicity of infection (MOI)

of 10. For the miRNA over-expression studies, the infectivity was determined by copGFP

expression as analyzed on FACScan (BD Bioscience). The copGFP positive cell populations

were isolated by fluorescence activated cell sorting (FACS) Vantage SE cell sorter (Becton

Dickinson) followed by plating on feeder cell layer for long-term culturing.

Embryoid body formation. hESC colonies were dispersed into cell aggregates containing

approximately 500 to 1,000 cells using 1 mg/mL collagenase IV. The cell aggregates were

then suspension-cultured in ultra-low attachment cell culture dishes with hESC differentiation

medium (without basic fibroblast growth factor): 80% Dulbecco‟s modified Eagle‟s

medium/F12, 1 mM L-glutamine, 0.1 mM β-mercaptoethanol, 0.1 mN non-essential amino

acids, 20% Knockout Serum Replacement. Culture media was changed every two days, for a

total of 6 days.

Cell samples collection and RNA preparation. Using the miRNeasy Mini Kit (Qiagen Inc.,

Valencia, CA), total RNA containing miRNAs was isolated separately from biological

duplicates of H7 hESCs and H7 hESC-CMs for microarray profiling. As positive control, total

RNA was isolated from left ventricular tissue derived from two 20-week old human fetus

hearts. As negative control, total RNA was isolated from IMR90 fetal fibroblasts (ATCC,

Manassas, VA). We prepared a total of eight distinct RNA samples for miRNA profiling. For

57

embryoid body experiments, total RNA was isolated from biological triplicates (each

experiment repeated minimum three times) of H7 hESCs and H7 embryoid bodies (day 6).

Total RNA concentration and purity were analyzed by spectrophotometry.

RT-PCR analysis of embryonic and cardiac specific transcriptions. The expression of

human embryonic cell markers (OCT4, NANOG, REX1) and cardiac-specific markers (ANF,

MEF2C, GATA4) were compared before and after hESC differentiation. 18S was used as

housekeeping gene control. The primer sets used in the amplification reaction are as follows:

Human OCT4 forward primer: 5‟-GGAAGGTATTCAGCCAAACGACCA-3‟

Human OCT4 reverse primer: 5‟-CTCACTCGGTTCTCGATACTGGTT-3‟

Human NANOG forward primer: 5‟-ACCAGAACTGTGTTCTCTTCCACC-3‟

Human NANOG reverse primer: 5‟-CCATTGCTATTCTTCGGCCAGTTG-3‟

Human REX1 forward primer: 5‟-GCGTACGCAAATTAAAGTCCAGA-3‟

Human REX1 reverse primer: 5‟-ATCCTAAACAGCTCGCAGAAT-3‟

Human ANF forward primer: 5‟-TAGGGACAGACTGCAAGAGG-3‟

Human ANF reverse primer: 5‟-CGAGGAAGTCACCATCAAACCAC-3‟

Human GATA4 forward primer: 5‟-TCAAATTGGGATTTTCCGGA-3‟

Human GATA4 reverse primer: 5‟- GCACGTAGACTGGCGAGGA-3‟

Human 18S forward primer: 5‟-ACACGGACAGGATTGACAGA-3‟

Human 18S reverse primer: 5‟-GGACATCTAAGGGCATCACAG-3‟

MicroRNA microarray construction, hybridization and scanning. Microarray assay was

performed using a service provider (LC Sciences, Houston, TX). The assay started with 4 to 8

µg total RNA sample, which was size fractionated using a YM-100 Microcon centrifugal filter

(Millipore, Billerica, MA) and the small RNAs (< 300 nt) isolated were 3‟-extended with a

poly(A) tail using poly(A) polymerase. An oligonucleotide tag was then ligated to the poly(A)

tail for later fluorescent dye staining. Hybridization was performed overnight on a µParaflo

microfluidic chip using a micro-circulation pump (Atactic Technologies, Houston, TX)163

. On

the microfluidic chip, each detection probe consisted of a chemically modified nucleotide

coding segment complementary to target microRNA (from miRBase,

58

http://microrna.sanger.ac.uk/sequences/) or other control RNA, and a spacer segment of

polyethylene glycol to extend the coding segment away from the substrate. The detection

probes were made by in situ synthesis using photogenerated reagent chemistry. Hybridization

used 100 µL 6xSSPE buffer (0.90 M NaCl, 60 mM Na2HPO4, 6 mM EDTA, pH 6.8)

containing 25% formamide at 34 °C. After RNA hybridization, tag-conjugating Cy3 or Cy5

dyes were circulated through the microfluidic chip for dye staining. Fluorescence images were

collected using a laser scanner (GenePix 4000B, Molecular Devices, Sunnyvale, CA) and

digitized using Array-Pro image analysis software (Media Cybernetics, Bethesda, MD).

MicroRNA profiling data analysis. Data were analyzed by first subtracting the background

and then normalizing the signals using a LOWESS filter (Locally-weighted Regression)164

.

Background is determined using a regression-based background mapping method. The

regression is performed on 5% to 25% of the lowest intensity data points excluding blank

spots. Raw data matrix is then subtracted by the background matrix. miRNA microarray data

were filtered to remove control probes, then log-transformed and scaled so that the mean and

standard deviation of the log2(miRNA expression) for each sample were 0 and 1, respectively.

For principal component analysis (PCA), the samples were plotted in the space of the first

three principal components. Samples and miRNAs were clustered based on Euclidean distance

using the complete linkage agglomeration method for the heat maps in Figure 13A,B. The

sample coordinates along each axis were scaled by the square-root of the variance in the

corresponding principal component to accurately reflect the relationship between the samples.

These analyses were performed using the R and Bioconductor statistical packages (versions

2.8.1 and 2.3, respectively).

59

Gene Set Enrichment Analysis (GSEA). GSEA was performed using the GeneSpring GX

10.0 software (Agilent, Technologies, Santa Clara, CA) and miRNA target gene sets

downloaded from TargetScan (www.targetscan.org/), and also a custom list of hESC markers.

Gene sets were considered significant with Q-value <0.25, as recommended50

. Briefly, the

primary result of GSEA is the enrichment score (ES), which reflects the degree to which a

gene set is overrepresented at the top or bottom of a ranked list of genes. The normalized

enrichment score (NES) is the primary statistic for examining gene set enrichment results. By

normalizing the enrichment score, GSEA accounts for differences in gene set size and in

correlations between gene sets and the expression dataset.

Ingenuity Pathway Analysis (IPA). Significant gene lists were generated from GeneSpring

software and uploaded to IPA for analysis. IPA software assigns biological functions to genes

using the Ingenuity Pathways Knowledge Base (Ingenuity Systems, Inc., Redwood City, CA).

This information is used to form networks to create an „interactome‟ of genes all involved in

specific biological processes.

Functional Analysis. The Functional Analysis identified the biological functions and/or

diseases that were most significant to the data set. Molecules from the dataset that met the P-

value cutoff of 0.05 and fold change cutoff of 1.4 were then associated with biological

functions and/or diseases in Ingenuity‟s Knowledge Base were. Right-tailed Fisher‟s exact test

was used to calculate a P-value determining the probability that each biological function

and/or disease assigned to that data set is due to chance alone.

Canonical Pathway Analysis. Canonical pathways analysis identified the pathways from the

Ingenuity Pathways Analysis library of canonical pathways that were most significant to the

data set. The significance of the association between the data set and the canonical pathway

was measured in 2 ways: 1) A ratio of the number of molecules from the data set that map to

60

the pathway divided by the total number of molecules that map to the canonical pathway is

displayed. 2) Fisher‟s exact test was used to calculate a P-value determining the probability

that the association between the genes in the dataset and the canonical pathway is explained by

chance alone.

Real-time quantitative PCR (qPCR). For mRNA qPCR, 2 µg of total RNA from each

sample was reversed transcribed with Superscript III (Invitrogen). For each sample, qRT-PCR

was performed in triplicate on an ABI 7900HT instrument (Applied Biosystems) using

Taqman primer probe sets (Applied Biosystems) for each gene of interest and a GAPDH

endogenous control primer probe set for normalization. Representative results are shown as

fold expression relative to undifferentiated hESCs unless otherwise stated. Error bars reflect

one standard deviation from the mean of three biological replicates unless otherwise stated.

For miRNA qPCR, single-stranded cDNA was synthesized from total RNA samples

using the TaqMan® MicroRNA Reverse Transcription Kit (Applied Biosystems, Foster City,

CA) according to the manufacturer‟s instructions. cDNA was then PCR amplified using

Human TaqMan® MicroRNA Assays (Applied Biosystems) for the following human miRNA:

miR-1, miR-302c, miR-302d, and miR-133a. Because miR-16 has been reported to be stably

expressed across human tissues165

, we also performed real-time quantitative PCR of this

microRNA to confirm its stable expression. Data were normalized to the endogenous control

genes RNU43 and U47. Results are shown as fold expression relative to undifferentiated

hESCs unless otherwise noted. Error bars indicate one standard deviation from the mean of

three technical replicates unless otherwise stated.

Electrophysiological characterization. Spontaneously beating clusters were microsurgically

dissected from hESC-derived EBs 14–21 days after initiating differentiation. After collagenase

61

dissociation and plate attachment, electrophysiological experiments of isolated single hESC-

CMs were performed using the whole-cell patch-clamp technique with an Axopatch 200B

amplifier and the pClamp9.2 software (Axon Instruments Inc., Foster City, CA). A xenon arc-

lamp was used to view GFP fluorescence at 488/530 nm (excitation/emission). Patch pipettes

were prepared from 1.5 mm thin-walled borosilicate glass tubes using a Sutter micropipette

puller P-97 and had typical resistances of 4-6 MΏ when filled with an internal solution

containing (mM): 110 K+ aspartate, 20 KCl, 1 MgCl2, 0.1 Na-GTP, 5 Mg-ATP, 5 Na2-

phospocreatine, 1 EGTA, 10 HEPES, pH adjusted to 7.3 with KOH. The external Tyrode‟s

bath solution consisted of (mM): 140 NaCl, 5 KCl, 1 CaCl2, 1 MgCl2, 10 glucose, 10 HEPES,

pH adjusted to 7.4 with NaOH. Current-clamp recordings were performed at 37 °C within 4 to

7 days after Lenti-miR™ transduction.

Statistical analysis. Non-microarray data are presented as mean S.D. Data were compared

using standard or repeated measures, using ANOVA where appropriate. Pairwise comparisons

were performed using a two-tailed Student‟s t–test. Differences were considered significant

for P-values<0.05.

RESULTS

To determine the miRNA-omes of hESC-CMs, cardiomyocytes were differentiated

from H7 hESCs based on a published protocol100

. Flow cytometry revealed greater than 90%

of cells were positive for the cardiac marker α-sarcomeric actinin (Fig. 11A). hESC-CMs

spontaneously beat in culture and exhibited a muscle-like morphology that was similar to

isolated left ventricular human fetal heart cells (Fig. 11B). While hESCs stained positive for

OCT4 (Fig. 11C), hESC-CMs stained positive for the cardiac specific markers Troponin-T,

MEF2C, and Connexin 43 (Fig. 11D). PCR further revealed that hESC-CMs express cardiac

62

63

Figure 11. Differentiation of hESCs to cardiomyocytes that express lineage-specific

genes. (A) Flow cytometry of differentiated hESC-CMs shows greater than 90% of cells are

positive for the cardiac marker α-sarcomeric actinin compared to isotype control. (B) hESC-

CMs spontaneously beat in culture and exhibited a muscle-like morphology that was similar

to 20-week left ventricular fetal heart (FH) cells. (Scale bars, 40 µm) (C) Undifferentiated

hESCs immunostain positive for the stem cell marker OCT4 (red). (Scale bars, 100 µm) (D)

hESC-CMs express cardiac specific markers MEF2C, cardiac Troponin-T (TropT), and

Connexin 43 (CX43) (Scale bars for 100X, 200X, and 630X are 200 µm, 60 µm, and 20

µm, respectively). (E) RT-PCR analysis reveals differences in expression of embryonic

(OCT4, NANOG, REX1) and cardiac markers (ANF, GATA4) in hESCs, hESC-CMs, fetal

heart tissue, and fetal fibroblasts. (F) qPCR of NKX2.5, MEF2C and GATA4 (early cardiac

genes) expression showed significant upregulation in hESC-CMs and fetal heart relative to

hESCs. Three cardiac myosin heavy chains were also upregulated: αMHC (MYH6), βMHC

(MYH7), and βMHC7b (MYH7B). Error bars represent one standard deviation from the

mean of biological triplicate experiments. (NS = Not Significant).

genes (NKX2.5, ANF, GATA4, αMHC (MYH6), βMHC (MYH7), and βMHC7b (MYH7B)),

with a corresponding decrease in embryonic genes (OCT4, NANOG, REX1) (Fig. 11E,F). By

quantitative PCR (qPCR), we noted similar, though not identical, gene expression levels in

hESC-CMs as compared to fetal heart (Fig. 11F), suggesting that these cells are different from

fetal heart, as has been previously described20

.

Next, we analyzed 711 unique miRNAs in hESC-CMs and human fetal left ventricles

using the Sanger miRBase Version 10.0 miRNA expression microarrays (LC Sciences,

Houston, TX). miRNA profiles for H7 hESCs and human fetal fibroblasts from our previous

publication24

were also included in the analysis. We first performed principal component

analysis (PCA), which is a reductionist method that provides an overview of the degree of

variation between samples (Figure 12). PCA indicated that the miRNA expression profile is

distinct for each of the four samples groups, with fetal heart and hESC-CMs exhibiting the

closest distance – and thus the most similar miRNA profiles, and hESCs the furthest from the

other three groups. Using clustered heat maps of subsets of miRNAs, we organized miRNAs

into two broad categories: miRNAs significantly expressed in hESCs (Fig. 13A) and were

64

similar to previous studies16, 146

, and miRNAs significantly expressed in the cardiac phenotype

(heart and hESC-CMs) (Figure 13B). MiRNA subsets are further refined in Fig. 13C-F. Some

miRNAs increased during differentiation (Fig. 13D) but were not cardiac-specific since they

were also expressed in fibroblasts, while other miRNAs decreased during differentiation (Fig.

13E). Of interest in this former group is miR-145, which has been previously reported to be

upregulated during differentiation and is an inhibitor of OCT4, SOX2, and KLF4143

. The

expression levels of a select group of miRNAs were confirmed with qPCR (Fig. 14).

We were of course particularly interested in miRNAs that appeared to be cardiac-

specific since these miRNAs may be critical for driving cardiac differentiation. Known

cardiovascular-related miRNAs include miRNAs -1, -133, and -208, which have roles in heart

failure, hypertrophy, stress response, cardiac and skeletal muscle development, and

differentiation156, 157, 160-162

. In our profiles, these miRNAs show robust expression in hESC-

Figure 12. PCA plots of the miRNA-ome data for hESC-CMs. (A) Graph showing the

fraction of total variance in the data that is explained by each of the eight principal

components. (B) Scatter plot of the samples in the space of the first three principal

components, showing the relationship between the miRNA profiles of the samples.

65

Figure 13. Subsets of miRNAs that are unique to hESCs and hESC-CMs. (A,B) Heat

maps of the miRNA microarray data, showing the normalized log2 miRNA expression level

for each miRNA (rows) in each sample (columns), according to the indicated color scale. (A)

miRNAs more highly expressed in hESCs. (B) miRNAs primarily expressed in cardiac

samples. (C-F) Microarray signal intensity plots of selected miRNAs. Signal values are the

mean of each pair of biological duplicate experiments; signal intensities below “threshold”

may not be quantitative. (c) miRNAs expressed highly in hESCs but not detectable in fetal

cell types. (D) miRNAs with increased expression during differentiation and in fetal cells.

(E) miRNAs with decreasing expression during differentiation. (F) Selected miRNAs with

high expression in cardiac cells (hESC-CM and/or fetal heart) but near or below threshold of

detection in hESCs and fibroblasts.

66

CMs and (with the exception of miR-208a) fetal heart, but are nearly undetectable in hESCs

and fibroblasts (Fig. 13F). Interestingly, we observed co-expression of miRNAs and their host

genes166

. For example, miR-208a is transcribed from an intron of the myosin heavy chain

αMHC (MYH6), and its expression therefore follows that of αMHC: higher expression in

hESC-CMs relative to fetal heart (see Fig. 11F). In contrast, miR-208b is transcribed from

intronic regions of βMHC (MYH7), the

predominant myosin heavy chain

isoform found in mature human cardiac

ventricles167

. miR-208b thus follows

βMHC‟s expression pattern: mildly

higher expression in fetal heart

compared to hESC-CMs, suggesting that

hESC-CMs may not be as

developmentally mature as fetal heart.

Of the cardiac phenotype-

associated miRNAs identified in Fig.

13B, we observed a novel miRNA, miR-

499-5p, that showed statistically

significant expression in the cardiac phenotypes but was undetectable in hESCs and

fibroblasts (Fig. 13F), a pattern similar to miR -1, -133 and -208b. This miRNA has been

previously reported to be important in human fetal cardiomyocyte differentiation168

and

regulation of murine muscle myofibers169

. Like the miR-208 family, miR-499 is located in an

intronic region of a myosin heavy chain gene, βMHC7B, and its trend in expression follows

that of βMHC7B: higher expression in fetal heart compared to hESC-CMs. Using TargetScan

(www.targetscan.org), an online database of predicted biological targets of miRNAs based on

Figure 14. qPCR confirms the transcriptional

changes of selected miRNAs observed in the

microarray data. Taqman qPCR of miR-1, -

133a, -302c, -302d, and -16. We included miR-

16 because it has been reported to exhibit

relatively stable expression across multiple

tissue types, though we did observe some

upregulation in fetal hearts. Data were

normalized to the endogenous control genes,

RNU48 and U47. Error bars represent one

standard deviation from the mean.

67

the presence of conserved 8mer and 7mer sites that match the seed region of each miRNA, we

found that miR-499 shared many targets with miR-208, and had some overlap with miR-1

targets (Fig. 15A). Canonical pathway analysis of the predicted targets for miR-499 using

Ingenuity Pathways Analysis (Ingenuity® Systems, www.ingenuity.com) revealed that many

targeted genes are involved in early cardiogenesis (e.g., the transcription factor GATA4121

)

and embryogenesis; targeted canonical pathways included WNT/β-catenin signaling170

, TGF-β

signaling, and cell cycle regulation (see Table 5 for selected pathways). These findings

suggest that miR-499 may be involved in maturation of the cardiomyocyte via inhibition of

embryogenesis and cardiogenesis pathways.

Figure 15. Target analysis of miR-499, miR-1, and miR-208. (A) Venn diagram of

TargetScan predicted gene targets. (B) Gene Set Enrichment Analysis (GSEA) of the

predicted targets for the three miRNAs, as well as the miR-302 cluster and a custom set of

26 pluripotentcy-related genes, in relation to published microarray data for hESCs, beating

EBs, hESC-CMs and fetal heart (GSE13834). The Normalized Enrichment Score indicates

degree of gene set enrichment in the indicated cell type relative to hESCs; >1 defines

upregulation, <1 defines downregulation. During cardiac differentiation, the gene targets of

miR-1, -208, and -499 are increasingly less enriched (relative to hESCs) due to higher

expression of these miRNAs in cardiac cell types. In contrast, the targets of the hESC-

specific miR-302 cluster are consistently highly enriched in all cell types relative to hESCs,

whereas pluripotency genes are consistently less enriched, as expected. Gene sets were

considered significant with Q-value <0.25 (see Materials and Methods).

68

Table 5: Selected canonical pathways predicted to be targets of miR-499

Ingenuity Canonical Pathways P-value Ratio Genes

Wnt/β-catenin Signaling 0.00002 5.95E-02

FZD8, TGFBR1, FZD4, CSNK1G1,

SOX6, LRP6, PPP2R2C, ACVR2B, APC,

SOX5

Factors Promoting Cardiogenesis in

Vertebrates 0.00004 7.87E-02

FZD8, TGFBR1, FZD4, LRP6,

ACVR2B, GATA4, APC

GM-CSF Signaling 0.00575 5.97E-02 ETS1, PPP3CB, SOS2, PPP3CA

Cell Cycle Regulation by BTG

Family Proteins 0.00759 8.33E-02 PPP2R2C, E2F3, CCRN4L

Role of NANOG in Mammalian

Embryonic Stem Cell Pluripotency 0.00851 4.39E-02 FZD8, FZD4, SOS2, GATA4, APC

Axonal Guidance Signaling 0.00955 2.48E-02

FZD8, FZD4, PPP3CB, SDC2, EFNB1,

SOS2, DCC, SRGAP2, PPP3CA,

PRKAR1A

Tight Junction Signaling 0.00977 3.57E-02 CPSF2, MYH10, CLDN11, TGFBR1,

PPP2R2C, PRKAR1A

Protein Kinase A Signaling 0.01905 2.53E-02 MYH10, TGFBR1, PPP3CB, DCC,

H3F3B, PPP3CA, CDC27, PRKAR1A

B Cell Receptor Signaling 0.02455 3.25E-02 ETS1, PPP3CB, VAV3, SOS2, PPP3CA

T Cell Receptor Signaling 0.02455 3.74E-02 PPP3CB, VAV3, SOS2, PPP3CA

PPARα/RXRα Activation 0.04467 2.73E-02 TGFBR1, SOS2, PRKAA2, ACVR2B,

PRKAR1A

Role of NFAT in Regulation of the

Immune Response 0.04571 2.56E-02

CSNK1G1, PPP3CB, SOS2, GATA4,

PPP3CA

Calcium Signaling 0.04898 2.43E-02 MYH10, PPP3CB, SLC8A3, PPP3CA,

PRKAR1A

Human Embryonic Stem Cell

Pluripotency 0.05129 2.7E-02 FZD8, TGFBR1, FZD4, APC

ERK/MAPK Signaling 0.05370 2.6E-02 ETS1, SOS2, H3F3B, PPP2R2C,

PRKAR1A

TGF-β Signaling 0.06026 3.61E-02 TGFBR1, SOS2, ACVR2B

Actin Cytoskeleton Signaling 0.09120 2.13E-02 MYH10, PIP5K1C, VAV3, SOS2, APC

GABA Receptor Signaling 0.10000 3.64E-02 ABAT, GABRA1

69

Next, we wanted to know whether the predicted targets of miR-499, along with miR-1

and miR-208, were significantly downregulated during cardiac differentiation as these

miRNAs become more highly expressed. For example, are the target genes of miR-499

downregulated in fetal heart relative to hESCs due to the higher expression of miR-499?

Using microarray data from our previous study20

(available on Gene Expression Omnibus,

GSE13834), we used Gene Set Enrichment Analysis (GSEA)50

to analyze the enrichment of

predicted miRNA targets in three successive stages of cardiac differentiation relative to

hESCs: beating embryoid body (EB, an intermediate stage of cardiac differentiation), hESC-

CM, and fetal heart. In Fig. 15B, the y-axis represents the normalized enrichment score

(NES), which is the primary statistic for examining GSEA results; NES>1 means the targets

are more highly expressed in the cardiac cell type, and NES<1 means the targets are more

highly expressed in hESCs. As expected, a custom set of 26 genes that are each associated

with pluripotency was more highly enriched in undifferentiated hESCs compared to the three

cardiac differentiation stages. In contrast, the predicted targets of the miR-302 cluster, which

was significantly upregulated in hESCs and has been previously associated with pluripotent

cell types24

, had the opposite pattern of expression: higher enrichment in the three stages of

cardiac differentiation relative to hESCs, as expected for an “embryonic” miRNA.

Importantly, the predicted target genes for miR-1, -208, and -499 each exhibited similar

enrichment patterns wherein the target genes were gradually reduced in expression during

cardiac differentiation, thus underscoring the similarity of miR-499 with the better

characterized miR-1 and miR-208 miRNAs.

We next wanted to determine how over-expression of miR-1 and miR-499 might

affect cardiac gene expression in differentiating hESCs. Using Lenti-miR™ MicroRNA

Precursor lentiviral constructs (System Biosciences, Mountain View, CA, Fig. 16A), we

stably transduced H7 hESCs with miR-1 or miR-499 (Fig. 17A). qPCR confirmed

70

upregulation of the two miRNAs (Fig. 16B). Cells were differentiated into embryoid bodies

(EBs) and qPCR then used to determine embryonic, mesodermal, and cardiac gene changes

(Fig. 17B). We noted significant upregulation of MEF2C in miR-499

EBs relative to miR-1

EBs and

WTEBs. Knockdown of miR-499 using a miRZip anti-miR-499 microRNA construct (System

Biosciences) did not upregulate MEF2C. MEF2C is required for activation of cardiac

contractile genes, and for cardiac structural development82

, suggesting that miR-499 may be

an important factor in differentiating hESCs. This is further evidenced by upregulation of

βMHC in miR-499

EBs relative to WT

EBs. miR-1

EBs exhibited upregulation of all three myosin

heavy chain isoforms. Interestingly, GATA4 (a predicted target of miR-499) was

downregulated in miR-499

EBs, as expected, but was significantly upregulated in miR-1

EBs,

suggesting that miR-1 and miR-499 each have distinct roles in cell fate biology. Clearly these

results highlight the ability of specific miRNAs such as miR-499 and miR-1 to alter the

expression of genes involved in the cardiac specification of hESCs.

Two studies have reported cardiac conduction abnormalities in murine hearts either

after transfection with miR-1171

or in mutants that lack miR-1 expression160

. To further explore

the functional effects of miR-1 and miR-499, we also transduced post-differentiated hESC-

Figure 16. Lentiviral vector for miRNA over-expression in hESCs. (A) Map of

microRNA precursor lentiviral construct (System Biosciences, Mountain View, CA). (B)

qPCR confirmation of upregulation of miRNA in transduced hESCs (WT = wild type). Error

bars represent one standard deviation from the mean.

71

Figure 17. Over-expression of miR-1 and miR-499 during hESC differentiation. (A)

miR-1 and miR-499 transduced hESCs are positive for GFP (scale bars, 100 µm), and

remain positive after differentiation to day 6 embryoid bodies (scale bars, 400 µm); wild-

type (WT) hESCs remain negative. (B) qPCR shows different patterns of gene expression

induced by the two miRNAs, as well as knockdown of miR-499 (anti-miR-499). Error bars

represent one standard deviation from the mean of biological quadruplicate experiments.

(*P<0.05 vs. WT

EB).

72

CMs with each miRNA. miR-1 transduction statistically decreased the beating rates of

spontaneously-contracting EB clusters, although miR-499 had no effect (6 independent

batches, P<0.05; Fig. 18A). Consistently, the pacemaker current (If), which is known to play a

major role in phase 4 depolarization, decreased its amplitude with miR-1 over-expression

(Fig. 18B). This finding confirms a previous report in which miR-1 was shown to decrease If

by targeting the pacemaker genes HCN2 and HCN4 in hypertrophic cardiomyocytes172

.

Figure 18. Electrophysiological changes in hESC-CMs after over-expression of miR-1

and miR-499. (A) Beating rates of spontaneously-contracting hESC-derived EB clusters

after miR-1 and miR-499 transduction (N=6). (B) Left: Representative If tracings recorded

from isolated single hESC-CMs with or without miR-1 transduction. Right: Bar graph

summarizing the If current. *P<0.05.

73

DISCUSSION

This study seeks to generate a systems-level overview of the miRNA profiles in

hESCs, hESC-CMs, and human fetal heart. Our results indicate that these profiles are highly

dynamic and specific during cardiac differentiation of hESCs. Several interesting miRNAs,

including miR-1, miR-208 and miR-133, that have been previously reported to be important in

cardiac development and disease, exhibited significant changes during cardiac differentiation.

Other miRNAs such as miR-499 that have poorly defined roles in development and

differentiation also showed cardiac-specific expression. In silico analysis of the predicted

targets for selected miRNAs revealed significant overlap between miR-208 and miR-499, both

of which are transcribed from the introns of myosin heavy chains.

Admittedly, the field of computational predictions of miRNA targets is still in its

infancy, and so we cannot assume complete confidence in their accuracy until further

validation through functional assays is performed. However, the analysis of target gene sets by

GSEA, rather than by individual genes, lends increased statistical power to the in silico

analysis of target predictions. Using this algorithm, we have shown that miR-499 shares many

similarities with previously identified cardiac-specific miRNAs in terms of gene targets and

expression pattern. In general, the targets of miR-1, -208, and -499 exhibited very similar

pattern of expression: high enrichment in EBs, moderate enrichment in hESC-CMs (though

miR-1 targets were still highly expressed in hESC-CMs), and significantly reduced

enrichment in fetal heart. We were surprised by some of the continued target expression in

hESC-CMs, which may be due to the heterogeneity of the hESC-CMs used in Cao et al. in

which only 43% were positive for the cardiac marker cardiac Troponin T, potential biological

differences between hESC lines (H7 vs. H9), and also the poorly understood roles these

miRNAs play during cellular differentiation. This latter point is based on the observation that

some miRNAs are co-expressed with their target mRNAs173

, and thus may represent complex,

74

incoherent feedback loops150

that help fine-tune the expression levels of important

transcription factors.

Lastly, with the ultimate goal of discovering factors that might enhance cardiac

differentiation, we selected miR-499 and miR-1 for functional studies. Each miRNA had

significant, though different, positive effects on cardiac differentiation, suggesting a powerful

role for miRNAs in influencing hESC fate decisions. What is likely necessary for efficient

differentiation of hESCs to cardiomyocytes is the simultaneous expression of miR-1, miR-

499, and others such as miR-208, that together regulate that molecular programs of hESC-

CMs. In the future, we believe miRNAs will play a key role in achieving higher yields of

hESC-CMs that can then be used for transplantation studies and, ultimately, clinical therapies.

75

3. INDUCING THE PLURIPOTENT STATE

MicroRNA profiling of human induced pluripotent stem cells24

.

ABSTRACT

MicroRNAs (miRNAs) are a newly discovered endogenous class of small noncoding RNAs

that play important posttranscriptional regulatory roles by targeting mRNAs for cleavage or

translational repression. Accumulating evidence now supports the importance of miRNAs for

human embryonic stem cell (hESC) self-renewal, pluripotency, and differentiation. However,

with respect to induced pluripotent stem cells (iPSC), in which embryonic-like cells are

reprogrammed from adult cells using defined factors, the role of miRNAs during

reprogramming has not been well characterized. Determining the miRNAs that are associated

with reprogramming should yield significant insight into the specific miRNA expression

patterns that are required for pluripotency. To address this lack of knowledge, we use miRNA

microarrays to compare the “microRNA-omes” of human iPSCs, hESCs, and fetal fibroblasts.

We confirm the presence of a signature group of miRNAs that is upregulated in both iPSCs

and hESCs, such as the miR-302 and 17-92 clusters. We also highlight differences between

the two pluripotent cell types, as in expression of the miR-371/372/373 cluster. In addition to

histone modifications, promoter methylation, transcription factors and other regulatory control

elements, we believe these miRNA signatures of pluripotent cells likely represent another

layer of regulatory control over cell fate decisions, and should prove important for the cellular

reprogramming field.

76

INTRODUCTION

Human embryonic stem cells (hESCs) have gained popularity as a potentially ideal

cell candidate for regenerative medicine. First isolated by James Thomson and colleagues in

19981, hESCs are derived from the inner cell mass of the human blastocyte and can be kept in

an undifferentiated, self-renewing state indefinitely. In contrast to adult stem cells, hESCs

have the advantage of being pluripotent, which endows them with the ability to differentiate

into virtually every cell type in the human body. However, the use of human embryos is

controversial in the US, and potential tissue rejection following transplantation in patients

remains problematic2.

One way to circumvent these issues is to generate induced pluripotent stem cells

(iPSCs). Mouse and human cells can be reprogrammed to pluripotency through ectopic

expression of defined transciption factors3-5, 174-179

. The first successful reprogramming of

human fibroblast cells into iPSCs was reported independently by Shinya Yamanaka (using

OCT4, SOX2, KLF4, c-MYC)4 and James Thomson (using OCT4, SOX2, NANOG, LIN28)

5.

The main advantage of this approach is that it does not need human embryos or oocytes to

generate patient-specific stem cells, and therefore can potentially bypass the ethical and

political debates that have surrounded this field for the past decade. Another important benefit

is that for the first time, disease-specific stem cells can be created, which will help scientists

understand the molecular mechanisms of many common inherited diseases6.

For a number of reasons, these reprogramming methods have so far been gradual and

slow, requiring weeks of cell culture with very low yield of iPSCs175, 179-181

. Inefficient

delivery of factors to the cells is certainly one obstacle, and this challenge is being actively

addressed by many groups. Another obstacle is the general lack of understanding of the

molecular changes that underlie reprogramming182, 183

. Understanding the molecular circuitry

of reprogramming will greatly benefit the field by providing new targets and pathways that

77

could increase the yield of iPSCs. Efforts to better integrate the genomic and epigenomic

networks that control reprogramming have been undertaken184

, but overall the specific

mechanisms required for more efficient reprogramming remain elusive.

One potential regulatory mechanism of reprogramming that has so far received little

attention is microRNAs (miRNAs). These small, noncoding RNAs play important

posttranscriptional regulatory roles by targeting messenger RNAs (mRNAs) for cleavage or

translational repression144

, and are key components of an evolutionarily conserved system of

RNA-based gene regulation in eukaryotes145

. Interestingly, hESCs are known to express

miRNAs that are often undetectable in adult organs such as miR-371, miR-372, miR-302a,

miR-302b, miR-302c, and miR-302d16, 17, 146, 154, 185

, whereas Dicer-deficient murine ESCs,

which cannot generate miRNAs, have been shown to be defective in differentiation147, 148

.

These and other studies suggest that miRNAs likely play key roles in human and murine ESC

gene regulation162, 185-189

. A recent study has attempted to incorporate miRNA gene regulation

into a model of transcriptional regulatory circuitry of ESCs by generating genome-wide maps

of binding sites for key ESC transcription factors such as OCT4, SOX2, and NANOG150

.

These ESC transcription factors were found to bind at many start sites of miRNA transcripts

that have been detected in ESCs, such as the miR-302 cluster. Clearly, at least a subset of

miRNAs seems to be involved in pluripotency, and these studies have contributed greatly to

the understanding of miRNA networks in ESCs.

While significant efforts are being applied to ESCs, no study has yet analyzed the

miRNA profile of human iPSCs. Determining the miRNAs that are associated with

reprogramming may yield significant insight into the specific miRNA expression patterns

needed for pluripotency. In this report, we use miRNA microarrays to compare the

“microRNA-omes” of human iPSCs, hESCs, and fetal fibroblasts. We confirm the presence of

a signature group of miRNAs that are upregulated in both iPSCs and hESCs, such as the miR-

78

302 and 17-92 clusters, as well as some subtle differences between the two pluripotent cell

types, including the miR-371/372/373 cluster. We also note the broad changes in miRNA

patterns between pluripotent cells and differentiated fibroblasts. These miRNA profiles are an

initial step towards a better understanding of the regulatory networks that govern pluripotency

and reprogramming.

MATERIALS AND METHODS

Immunohistochemical analysis. IMR90 fibroblasts (ATCC, Manassas, VA) were cultured

on 0.1% gelatin-coated coverslips, and hESCs (Wicell H7 line) and iPSCs (from Thomson

lab) on Matrigel-coated coverslips. Cells were fixed in 2% formaldehyde in PBS,

permeabilized with 1% tritonX-100 in PBS, and blocked with 5% bovine serum albumin in

PBS. Cells were then labeled with each embryonic stem cell marker plus appropriate

AlexaFluor secondary antibodies (Invitrogen) and DAPI (Vectorshield Laborary). Microscopy

was performed using a Leica DM-IL inverted microscope (Leica Microsystems USA,

Bannockburn, IL) and QImaging Retiga 2000R high-speed digital CCD camera (QImaging,

Burnaby, BC, Canada). QCapture Pro 5.1 (QImaging, Burnaby, BC, Canada) was used for

image acquisition.

Embryoid body formation. Human embryoid body formation was accomplished by treating

proliferating cells (iPSCs or hESCs) with collagenase IV and transferring them into

suspension culture in differentiation media consisting of DMEM-F12 Media (GIBCO)

supplemented with 20% FBS, L-Glutamine, β-mercaptoethanol, and non-essential amino

acids. Embryoid bodies were maintained in suspension culture for 14 days, changing the

media every two days.

79

Cell samples collection and RNA preparation. Using the miRNeasy Mini Kit (Qiagen Inc.,

Valencia, CA), total RNA containing miRNA was isolated separately from biological

duplicates of iPSCs, ESCs, IMR90 fetal fibroblasts. In total, we prepared six distinct RNA

samples for miRNA profiling. Total RNA concentration and purity were analyzed by

spectrophotometry.

RT-PCR analysis of embryonic and cardiac specific transcriptions. The expression of

human embryonic cell markers (OCT4, NANOG, SOX2, REX1, DNMT3B, GDF3), as well as

differentiation markers (Brachyury – mesoderm, AFP – endoderm, and Ncam1 –

neuroectoderm), was compared before across all three groups (in biological duplicates). 18S

was used as housekeeping gene control. The primer sets used in the amplification reaction are

as follows:

Human OCT4 forward primer: 5‟-GGAAGGTATTCAGCCAAACGACCA-3‟

Human OCT4 reverse primer: 5‟-CTCACTCGGTTCTCGATACTGGTT-3‟

Human NANOG forward primer: 5‟-ACCAGAACTGTGTTCTCTTCCACC-3‟

Human NANOG reverse primer: 5‟-CCATTGCTATTCTTCGGCCAGTTG-3‟

Human SOX2 forward primer: 5‟-GGGAAATGGGAGGGGTGCAAAAGAGG-3‟

Human SOX2 reverse primer: 5‟-TTGCGTGAGTGTGGATGGGATTGGTG-3‟

Human REX1-1 forward primer: 5‟-GCGTACGCAAATTAAAGTCCAGA-3‟

Human REX1 reverse primer: 5‟-ATCCTAAACAGCTCGCAGAAT-3‟

Human DNMT3B forward primer: 5‟-TGCTGC TCCAGGGCCCGATACTTC-3‟

Human DNMT3B reverse primer: 5‟-TCCTTTCGAGCTCAGTGCACCACAAAAC-3‟

Human GDF3 forward primer: 5‟-CTTATCTCGTAAAGGGCTGGG-3‟

Human GDF3 reverse primer: 5‟-GTGCCAACCCAGGTCCCGGAAGTT-3‟

Human Brachyury forward primer: 5‟-GCGGGAAAGAGCCTGCAGTA-3‟

Human Brachyury reverse primer: 5‟-TTCCCCGTTCACGTACTTCC-3‟

Human AFP forward primer: 5‟-GCTGGATTGTCTGCAGGATGGGGAA-3‟

Human AFP reverse primer: 5‟-TCCCCTGAAGAAAATTGGTTAAAAT-3‟

Human NCAM1 forward primer: 5‟-GCCAGGAGACAGAAACGAAG-3‟

Human NCAM1 reverse primer: 5‟-GGTGTTGGAAATGCTCTGGT-3‟

Human 18S forward primer: 5‟-ACACGGACAGGATTGACAGA-3‟

Human 18S reverse primer: 5‟-GGACATCTAAGGGCATCACAG-3‟

80

MicroRNA microarray construction, hybridization and scanning. Microarray assay was

performed using a service provider (LC Sciences, Houston, TX). The assay started with 4 to 8

µg total RNA sample, which was size fractionated using a YM-100 Microcon centrifugal filter

(Millipore, Billerica, MA), and the small RNAs (< 300 nt) isolated were 3‟-extended with a

poly(A) tail using poly(A) polymerase. An oligonucleotide tag was then ligated to the poly(A)

tail for later fluorescent dye staining. Hybridization was performed overnight on a µParaflo

microfluidic chip using a micro-circulation pump (Atactic Technologies, Houston, TX)163

. On

the microfluidic chip, each detection probe consisted of a chemically modified nucleotide

coding segment complementary to target microRNA (from miRBase,

http://microrna.sanger.ac.uk/sequences/) or other control RNA, and a spacer segment of

polyethylene glycol to extend the coding segment away from the substrate. The detection

probes were made by in situ synthesis using photogenerated reagent chemistry. Hybridization

used 100 µL 6xSSPE buffer (0.90 M NaCl, 60 mM Na2HPO4, 6 mM EDTA, pH 6.8)

containing 25% formamide at 34 °C. After RNA hybridization, tag-conjugating Cy3 or Cy5

dyes were circulated through the microfluidic chip for dye staining. Fluorescence images were

collected using a laser scanner (GenePix 4000B, Molecular Devices, Sunnyvale, CA) and

digitized using Array-Pro image analysis software (Media Cybernetics, Bethesda, MD).

MicroRNA microarray normalization and data analysis. Microarray normalization

removes system related variations, such as sample amount variations, different labeling dyes,

and signal gain differences of scanners so that biological variations can be faithfully revealed.

Data were analyzed by first subtracting the background and then normalizing the signals using

a LOWESS filter (Locally-weighted Regression)164

. Background is determined using a

regression-based background mapping method. The regression is performed on 5% to 25% of

the lowest intensity data points excluding blank spots. Raw data matrix is then subtracted by

81

the background matrix. Data adjustment includes data filtering, Log2 transformation, and gene

centering and normalization. The data filtering removes miRNAs with normalized intensity

values below a threshold value of 32 across all samples. Gene centering and normalization

transform the Log2 values using the mean and the standard deviation of individual genes

across all samples. T-values are calculated for each miRNA between groups, and P-values are

computed from the theoretical t-distribution. miRNAs with P-values below a critical P-value

(typically 0.01) are selected for cluster analysis. The clustering was done using hierarchical

methods, with average linkage and Euclidean distance metrices190

. The clustering plot was

generated using TIGR MeV (Multiple Experimental Viewer) software from The Institute for

Genomic Research. Principal component analysis was performed using the R statistical

package (cran.r-project.org) (Fig. 21). For Fig. 21A, in order to accurately represent the

distances between samples in principal component space, the loading of the samples along

each principal component was scaled (multiplied) by the standard deviation along that

component.

Quantitative RT-PCR. For mRNA qRT-PCR, 2 µg of total RNA from each sample was

reversed transcribed with Superscript III (Invitrogen). For each sample, qRT-PCR was

performed in triplicate on an ABI 7900HT instrument (Applied Biosystems) using Taqman

primer probe sets (Applied Biosystems) for each gene of interest and a GAPDH endogenous

control primer probe set for normalization. Representative results are shown as fold expression

relative to undifferentiated hESCs unless otherwise stated. Error bars reflect one standard

deviation from the mean of three technical replicates unless otherwise stated. miRNA qRT-

PCR was performed with miRNA Taqman Expression Assays (Applied Biosystems) and the

miRNA Reverse Transcription kit (Applied Biosystems). For each miRNA analyzed, 10 ng of

total RNA was reverse transcribed with a miRNA-specific primer. RNU48 was used as the

82

endogenous control. Results are shown as fold expression relative to undifferentiated hESCs

unless otherwise noted. Error bars indicate one standard deviation from the mean of three

technical replicates unless otherwise stated.

Statistical Analysis. All results are expressed as mean±SD. The Student‟s t test was used, and

P-values of <0.05 were considered to indicate significant differences between two groups.

RESULTS

Characterization of fibroblasts, iPSCs and hESCs. We obtained iPSCs from the

James Thomson lab (University of Wisconsin-Madison), which were originally derived from

IMR90 fetal fibroblasts (ATCC, Manassas, VA) using the reprogramming factors OCT4,

SOX2, NANOG, and LIN285. hESCs (H7 from Wicell, Madison, WI) and iPSCs were

maintained in the undifferentiated state using mTeSR™1 medium (StemCell Technologies,

Vancouver, Canada). iPSC colonies exhibited an embryonic-like morphology that was similar

to hESCs (Fig. 19). In contrast to IMR90 fibroblasts, iPSCs and hESCs stained positive for a

number of well-characterized embryonic markers. We also performed RT-PCR gene

expression analysis of selected embryonic and germ layer genes (Fig. 20A). Confirming our

immunostaining results, iPSCs and hESCs expressed the same set of embryonic genes (OCT4,

SOX2, REX1, DNMT3B, GDF3), but were negative for the differentiation markers Brachyury

(T) (mesoderm), AFP (endoderm), and NCAM1 (neuroectoderm). We further verified the

expression of OCT4, SOX2, and NANOG in iPSCs and hESCs using quantitative RT-PCR

(Fig. 20B). The relative expression levels of these pluripotency genes were similar between

the two pluripotent cell types. Taken together, these findings indicate that our iPSCs and

hESCs were ostensibly free of spontaneously-differentiated contaminants, and expressed

appropriate levels of genes associated with pluripotency.

83

MiRNA profiling. We next wanted to determine the miRNA profiles of all three cell

types: IMR90 fibroblast, iPSCs, and hESCs. Using Sanger miRBase Version 10.0 miRNA

expression microarrays (LC Sciences, Houston, TX), we analyzed 697 unique miRNAs across

biological duplicates of each cell type. We performed principal component analysis (PCA) on

the miRNA profiles to assess the overall similarities and differences in miRNA expression

between human iPSCs, hESCs, and fibroblasts. Of the six principal components (PC), the first

Figure 19. Morphology and pluripotent

marker staining of human fibroblasts,

iPSCs, and hESCs. (A) Fibroblasts

exhibit typical morphology of these cells,

do not grow in colonies, and do not stain

for common markers of embryonic cells.

(B) iPSCs exhibit colony formation and

show robust staining for common

embryonic markers. Note that mouse

embryonic fibroblasts (MEFs) were co-

cultured with iPSCs and so also stain

positive for DAPI. (C) Similar to iPSCs,

hESCs grow in colonies and stain positive

for the same group of embryonic markers.

84

Figure 20. PCR and microarray analysis of human fibroblasts, iPSCs, and hESCs.

mRNA was isolated separately from biological duplicates of IMR90 fetal fibroblasts,

iPSCs, and hESCs. (A) As expected, the embryonic genes (OCT4, NANOG, SOX2, REX1,

DNMT3B, GDF3) were detectable in iPSCs and hESCs, but not in fetal fibroblasts. To

ensure that our cells were in an undifferentiated state and had no differentiated

contaminants, we also analyzed markers of mesoderm (Brachyury, T), endoderm (AFP),

and neuroectoderm (Ncam1), which were all negative. Human 18S was used as loading

control. (B) Quantitative RT-PCR analysis of the pluripotency genes OCT4, SOX2 and

NANOG confirms similar expression levels of these genes in iPSCs and hESCs. Error bars

represent one standard deviation from the mean. (*P-value <0.05 vs. fibroblasts; NS = Not

Significant). (C) Pearson clustering of the microarray data for the three cell types (average

of duplicate experiments for each sample).

three explain 78%, 18%, and 2%, respectively, of the total variance in the data (Fig. 21A). All

three cell types had similar weights along the first PC, suggesting that this component

represents variation in miRNA levels due to factors that are independent of cell type (e.g.,

promoter strength, pre-miRNA stability) (Fig. 21B). The second and third PCs appear to

account for the cell-type specific differences in miRNA expression. The second PC

distinguishes fibroblasts from both iPSCs and hESCs, suggesting that it corresponds to the

85

pluripotency state of the cell. Finally, iPSCs and hESCs were separated, though to a lesser

degree, along the third PC, indicating that these two cell types are not identical in their

miRNA expression profiles. Thus

the first three principal components

indicated that the miRNA expression

profile of iPSCs is more similar to

that of hESCs than fibroblasts, but

still distinct from either.

Using clustered heat maps of

the miRNA datasets, we observed

many miRNAs that were highly

expressed in iPSCs and hESCs but

not in fibroblasts (Fig. 22A); a

smaller group of miRNAs that

exhibited dissimilar expression in

iPSCs and hESCs (Fig. 22B); and

miRNAs that were upregulated in

fibroblasts but downregulated in both

iPSCs and hESCs (Fig. 22C). In

addition to clustered heat maps, we

have also included normalized

intensity plots of selected miRNAs

that have previously been associated

with developmental or stem cell processes. These intensity plots give an indication of the

absolute, rather than relative, miRNA expression across the three cell populations. Lastly, we

Figure 21. PCA plot of the transcriptomes for

fibroblasts, iPSCs, and hESCs. (A) Barplot

showing the fraction of total variance in the miRNA

data explained by each of the six principal

components. The first three components together

account for 99% of the total variance. (B) The six

miRNA profiles, two each of human iPSCs (hiPS),

ESCs (hESC), and fibroblasts (hFib), are plotted

along the first three principal components. The

coordinates of each point indicate the relative weight

of each principal component in that profile. As

indicated by the scatter, the miRNA profile of iPSCs

appears to be more similar to that of hESCs than

fibroblasts, but still distinct from either.

86

87

Figure 22. Heat maps and signal intensity plots of miRNA expression across human

fibroblasts, iPSCs, and hESCs. ANOVA analysis demonstrates statistically significant

differential miRNA regulation across the three samples. miRNAs with P-values below 0.01

were selected for cluster analysis. (A) miRNAs upregulated in both iPSCs and hESCs

compared to fibroblasts. Highlighted are the normalized signal intensity plots for the miR-

302 cluster and miR-17-92 cluster, including its paralogs miR-106a-92 and miR-106b-25.

(B) miRNAs exhibiting opposite expression in iPSCs and hESCs. Note that the normalized

signal intensities of the miRNAs are not dramatic, though are statistically significant. In

contrast to the group of miRNAs upregulated in hESCs but not iPSCs (top panel), we

observed only one miRNA (miR-886-5p) that was upregulated in iPSCs but not hESCs

(bottom panel). (C) miRNAs upregulated in fibroblasts but downregulated in iPSCs and

hESCs. Highlighted in the top and bottom panels are selected miRNAs known to be

important in ESC biology and/or organismal development, such as the let-7 cluster that has a

well-established role in late development timing. The clustering was done using hierarchical

methods and was performed with average linkage and Euclidean distance metrices. The

“threshold” value denotes signal intensities less than 32, a cutoff used by the microarray

service provider below which quantitation of signal may be inaccurate.

confirmed the microarray expression patterns of selected miRNAs using quantitative RT-PCR,

and also verified that iPSCs could be differentiated into embryoid bodies that express similar

patterns of pluripotency genes as hESCs (Fig. 23).

The “pluripotent” miRNAs identified in Fig. 22A are similar to results from previous

studies of hESCs and embryonic tissues16, 146, 154, 191

. miRNAs are frequently transcribed

together as polycistronic primary transcripts that are then processed into multiple individual

mature miRNAs. The genomic organization of these miRNA clusters is often highly

conserved, suggesting an important role for coordinated regulation and function. The most

dramatic fluctuation in miRNA expression occurred in the miR-302 cluster, which has been

consistently associated with ESCs in numerous miRNA profiling and sequencing studies16, 17,

146, 154, 185. Quantitative RT-PCR of miR-302b, one of the miR-302 cluster members, shows

that it is similarly downregulated in both iPSCs and hESCs upon differentiation to day 14

embryoid bodies (Figure 23D). This group of eight miRNAs, which includes miR-367, is an

evolutionarily conserved cluster located in the antisense intron of the protein-coding gene

88

Figure 23. Quantitative RT-PCR of selected miRNAs confirms expression patterns seen

in the microarray data, and also reveals miRNA expression during differentiation. (A)

miR-17, miR-92a, and miR-302b show similarly high expression in both pluripotent cell

types, but not in fibroblasts. In contrast, miR-372 and miR-373 are expressed at lower levels

in iPSCs relative to hESCs, and at even lower levels in fibroblasts. (B) Also confirming the

microarray data, miR-886-5p is expressed at statistically significant higher levels in

fibroblasts and iPSCs compared to hESCs, and let-7a is more highly expressed in fibroblasts

relative to both pluripotent cell types. Expression of let-7a remains different between iPSCs

and hESCs however. (C) Differentiation of iPSCs and hESCs to day 14 embryoid bodies

reveals downregulation of the pluripotency genes OCT4 and NANOG, as expected. (D)

miR-302b, part of the miR-302 cluster that is highly expressed in pluripotent cells, is

downregulated in both iPSCs and hESCs at day 14 of differentiation. (E) miR-372 and miR-

373 are moderately upregulated or unchanged, respectively, with differentiation. (F) miR-

886-5p is downregulated in both pluripotent cell types during differentiation. Error bars

represent one standard deviation from the mean of biological duplicate experiments. (#P-

value <0.05 vs. fibroblasts; *P-value <0.01 vs. fibroblasts; NS = Not Significant).

89

LARP7, a member of the La ribonucleoprotein domain family that is involved in RNA

metabolism 192, 193

. This poorly understood cluster of miRNAs, along with the LARP7 gene,

may represent one important regulatory switch for the transition from embryonic to mature

phenotypes in hESCs.

Also apparent from Fig. 22A is the upregulation of the miR-17-92 cluster and its

paralogs in iPSCs and hESCs. These miRNAs are components of three paralogous clusters

including, miR-17-92 at 13q31.3, miR-106a-92 at Xq26.2, and miR-106b-25 at 7q22 with

extensive sequence homologies194

. In the human genome, the miR-17-92 cluster is located

within intron 3 of the C13orf25 gene, and encodes six miRNAs (miR-17, miR-18a, miR-19a,

miR-20a, miR-19b-1, and miR-92-1)195, 196

. Both the sequences of these mature miRNAs and

their organization are highly conserved in all vertebrates, and they have attracted attention due

to their possible oncogenicity196

. This cluster has also been implicated in mammalian

development, where loss-of-function of the miR-17-92 cluster resulted in mice that died

shortly after birth with lung hypoplasia and cardiac defects197

. Together with the miR-302

cluster and other upregulated miRNAs identified in our profiles, the miR-17-92 cluster likely

represents an important pluripotency regulatory element.

We were very interested to find any differences in miRNA expression between iPSCs

and hESCs, and in Figure 22B we highlight a small group of miRNAs that exhibit dissimilar

expression between the two pluripotent cell types. We observed a group of miRNAs that were

expressed more highly in hESCs than in iPSCs (top panel), and a single miRNA, miR-886-5p,

that was expressed more highly in iPSCs than in hESCs (bottom panel). To confirm these

microarray results, we used quantitative RT-PCR to measure the levels of miR-886-5p, miR-

372, and miR-373 (Fig. 23A,B); our results confirm that these miRNAs are expressed

differently in iPSCs and hESCs. As a follow-up experiment we also assessed the expression of

selected miRNAs after differentiation to day 14 embryoid bodies (Fig. 23E,F). In both

90

pluripotent cell types, the expression patterns of these miRNAs were similar upon

differentiation. Given the lack of annotation of these miRNAs, it is difficult to ascertain their

biological significance for reprogramming and pluripotency at present.

Interestingly, except for two miRNAs, miR-629 and miR-96, all of the miRNAs that

exhibited higher expression in hESCs relative to iPSCs (listed in the top panel of Fig. 22B) are

located together within a 130 kb intergenic region of chromosome 19. This enormous meta-

region of clustered miRNAs includes the miR-371/372/373 cluster, which has previously been

shown to be upregulated in hESCs16, 146, 154, 185

, as well as a larger 54 miRNA cluster spanning

96 kb198

, much of which has also been shown to be upregulated in hESCs17

. We were therefore

surprised to find that these same chromosome 19 clusters were not expressed at similar levels

in iPSCs. Based on this observation, one could hypothesize that robust expression of these

clustered miRNAs may not be critical for pluripotency, in contrast to a cluster such as miR-

290 (found in mouse) that controls de novo DNA methylation and is thus critical for cellular

pluripotency188

. In making this hypothesis, however, it is important to note that the normalized

signal intensities, a measure of absolute rather than relative expression, show that the overall

change in expression among these miRNAs is small. As comparison, the signal intensities for

the miR-302 cluster show a much more dramatic fluctuation in expression levels (Fig. 22A).

Therefore, though these changes are statistically significant, whether they are biologically

significant remains to be determined. It is safe to say that differences, albeit small, do exist

between the miRNA profiles of iPSCs and hESCs, and that some of these differences may

help define the specific miRNAs required for induction and maintenance of pluripotency.

Fig. 22C shows the miRNAs that are upregulated in fibroblasts with respect to iPSCs

and hESCs. Many of these miRNAs have little annotation and so have poorly defined roles in

development and pluripotency, though they can be assumed to be important for maintenance

of the differentiated or fetal fibroblast phenotype. The upper half of the magnified heat map in

91

Fig. 22C shows miRNAs that are minimally downregulated in iPSCs (shaded in black), but

that are significantly downregulated in hESCs. A number of these miRNAs have previously

been reported to be upregulated in differentiated cells relative to hESCs16, 17, 154

, and we have

plotted the intensity values for a few of these miRNAs in Fig. 22C. Though there is an

obvious downward trend in miRNAs when transitioning from the fibroblast to iPSC state, the

expression levels of many do not completely decrease to the levels observed in hESCs. One

example is let-7a, which showed statistically significant higher expression in iPSCs relative to

hESCs as confirmed by quantitative RT-PCR (Fig. 23B). These observations underscore our

general findings that there are subtle differences in the miRNA profiles between the two

pluripotent cell types.

One group of miRNAs in Fig. 22C that has been extensively studied is the let-7

cluster. This cluster is known to be expressed sequentially at specific stages of development to

help coordinate developmental timing145, 199

. One recent study has also shown that LIN28, one

of the reprogramming factors used by the James Thomson lab to induce pluripotency in

fibroblasts, blocks an early transcript of let-7g that is preferentially expressed in adult cells200

.

Furthermore, the promoters of both let-7g and LIN28 are occupied by the embryonic

transcription factors OCT4, SOX2, NANOG, and TCF3 in mice, suggesting that these factors

promote the transcription of both primary let-7g and LIN28, which then blocks the maturation

of let-7g150

. Taken together, these studies, along with our new findings, suggest one possible

mechanism for the induction of pluripotency in adult cells using ectopic transcription factor

expression.

DISCUSSION

Here we present the first profiling study of the human iPSC miRNA-ome. Our profiles

demonstrate a high degree of similarity between iPSCs and hESCs, though with some

92

differences. Because they can regulate numerous genes, often in common pathways, miRNAs

may be regulators of cellular processes, akin to transcription factors that control entire

programs of cellular differentiation and organogenesis. Therefore, in addition to histone

modifications, promoter methylation, transcription factors and other regulatory control

elements, these miRNA signatures likely represent another layer of regulatory control for cell

fate decisions, and should prove significant for the cellular reprogramming field.

The idea that miRNAs may contribute to the induction and maintenance of

pluripotency arises from their well established role in development and ESC biology. During

development, miRNAs are known to interfere with the expression of mRNAs encoding factors

that control developmental timing, stem cell maintenance, and other developmental and

physiological processes145

. However, though over 450 human miRNA have been described191

,

each of which is predicted to target tens if not hundreds of different mRNAs, only a fraction of

them have been annotated; the great majority of miRNAs have poorly-defined roles in cell fate

decisions. This limited annotation, combined with a lack of knowledge of the transcriptional

start sites, promoter regions, and downstream mRNA targets of miRNAs, has made it difficult

to comprehensively integrate miRNAs into a broader understanding of molecular networks.

Unlike other molecular mechanisms that have benefited from extensive research beginning in

the latter half of the 20th century, the relatively new field of miRNAs, especially in the case of

stem cells, has thus suffered from a lack of molecular context.

Given these issues, attempting to connect iPSC miRNA profiles to the molecular

landscapes that underlie reprogramming will raise a number of basic questions. What regulates

these small molecules, and what are their effects on downstream gene networks? If non-

coding RNAs are regulated similarly to transcription factors, what advantage(s) do they offer

over transcription factors in terms of regulation of pluripotency? Could miRNAs, each

potentially targeting hundreds of mRNAs, represent nodes of control that dictate a cell‟s

93

regulatory dynamics? These questions lead to a general observation regarding the fundamental

difference between miRNA and mRNA activity. By inhibiting the translation of mRNAs

within the cytoplasm, miRNAs directly affect the existing transcriptome of the cell. That is,

the pool of mRNA in the cytoplasm – the “old” mRNA pool - is the immediate target of

miRNAs. In contrast, transcription factors and other elements are responsible for activation

and suppression of genes that will ultimately become the “new” mRNA pool. In the case of

cellular reprogramming, as well as stem cell differentiation, it is therefore tempting to

hypothesize that miRNAs can help transition the old mRNA pool to the new, with each layer

of control required for efficient transformation of the cell. These questions and more will be

active areas of investigation in stem cell biology and beyond for the foreseeable future.

94

3. INDUCING THE PLURIPOTENT STATE

Current and future directions: Improving induction of pluripotency with

microRNAs.

INTRODUCTION

For iPSC generation, the current method of integrating transcription factor genes (e.g.

OCT4, SOX2, KLF4, c-MYC, LIN28, and NANOG) into the adult cell genome requires

weeks of cell culture, and has so far resulted in very low yield of iPSCs. Inefficient delivery of

factors to the cells is certainly one obstacle, but more fundamentally is the failure of the

introduced transcription factors to alter the pre-existing mRNA pool (the transcriptome) that is

actively being translated into protein (the proteome) within the adult cell. Thus, the adult cell‟s

existing transcriptome continues to maintain the adult phenotype while the pluripotent mRNA

slowly reaches critical mass and overwhelms the adult proteome with variable success; this

inefficient process results in delayed pluripotency and potentially incomplete reprogramming

(Fig. 24). To address these problems, we propose a radical new method for reprogramming

that is based on miRNAs. The primary advantage of miRNAs is that, unlike transcription

factors, they will directly and immediately alter the adult transcriptome, and by extension the

adult proteome, leading to increased efficiency for inducing pluripotency. This innovative

approach will create an entirely new paradigm for iPSC generation, and will also help

elucidate the molecular processes that underlie pluripotency and reprogamming.

MATERIALS AND METHODS

Recently, we have determined the miRNA signature of iPSCs that have been derived

using reprogramming transcription factors24

. In collaboration with System Biosciences

(Mountain View, CA), we created a custom “pool” of miRNA over-expression constructs that,

95

Figure 24. Rationale for using microRNAs in combination with transcription factors to

increase reprogramming efficiency. Transcription factors alone are inefficient in

overwhelming the existing proteome of the adult cell (e.g. fibroblast) during reprogramming

(middle of panel). In contrast, microRNAs inhibit translation of targeted regions of the

transcriptome, thus potentially expediting the transformation of an adult cell to a pluripotent

cell (bottom of panel).

96

in total, will over-express 28 miRNAs after stable integration into the host genome (Fig. 25A).

These miRNAs were chosen for evaluation as potential reprogramming factors based on their

high expression in pluripotent cells and minimal expression in differentiated cells, including

the miR-302 and miR-17-92 clusters24

. For our miRNA reprogramming experiments, we used

human adipose stem cells (hASCs) as the starting adult cell line27

. hASCs have a number of

advantages over other somatic cell types such as fibroblasts since they can be isolated in large

quantities (100 ml of human adipose tissue yields about 1×106 cells) with minimal

morbidity201

. On Day 20 post-infection, cells were stained with Alkaline Phosphatase (AP), a

common embryonic marker, and the number of positive clusters counted. The reprogramming

efficiency was then estimated by dividing the number of AP+ clusters on Day 20 by the

original number of plated cells on Day 0.

RESULTS AND DISCUSSION

Starting with freshly isolated hASCs from donor patients, we infected cultures with

the miRNA pool, the miR-302 cluster alone, or the miR-17-92 cluster alone. We also co-

infected cells with combinations of lentiviral Yamanaka factors (OCT4, SOX2, KLF4, c-

MYC) in addition to the miRNA vectors. Clearly there is a trend towards increased AP

staining when Yamanaka factors were combined with the miRNA pool or with the miR-302

cluster (Fig. 25B). The miR-17-92 cluster did not significantly increase the number of AP+

clusters, and may even have diminished AP+ staining when it was used in combination with

the four Yamanaka factors. After further culture and immunostaining for other embryonic

markers such as TRA-1-60, we confirmed a 2-3 fold increase in iPSC colony generation with

the miR-302 cluster + Yamanaka factor combination (compared to Yamanaka factors alone).

Interestingly, we have not been able to derive iPSCs using miRNAs alone. While this

work is continuing, it has become increasingly clear that the 28 miRNAs tested so far can only

97

generate iPSCs when they are used in combination with Yamanaka transcription factors. We

hypothesize that miRNAs‟ inability to induce pluripotency by themselves may be due to their

primary role in modulating gene expression, thus acting as genetic “buffers” of key

pluripotency proteins152

, rather than acting directly on DNA to inhibit or promote gene

Figure 25. Increased reprogramming efficiencies when using combinations of miRNAs

and transcription factors. (A) The custom miRNA pool, containing nine individual

constructs carrying a total of 28 miRNAs. Additionally, the miR-302 cluster and miR-17-92

cluster were each tested independently to determine their reprogramming ability. A map of

the miR-302abcd lentiviral vector is also shown. (B) Percentage Alkaline Phosphatase (AP)

positive staining on Day 20 after hASCs were infected with combinations of miRNA and

Yamanaka transcription factors. (O=OCT4, S=SOX2, K=KLF4, C=c-MYC). Experiment

was repeated twice.

98

transcription. Put another way, though they may directly fine-tune and even inhibit the

translation of mRNAs to proteins, they only indirectly promote or inhibit the expression of the

mRNAs themselves, making fundamental transformation of the cell impossible. However, we

do still observe some increase in reprogramming efficiency with the addition of miRNAs to

the reprogramming cocktail (Fig. 25B), so we are confident that they act as modulators of the

transition between cellular phenotypes. This is further substantiated by Robert Blelloch and

colleagues who have recently shown that miRNAs may be used in place of c-MYC for

inducing pluripotency in mouse cells202

. With the advent of next generation sequencing

technologies, increasing numbers of novel miRNAs are being discovered in embryonic stem

cells16, 17, 203

. Future work will need to assess their roles in pluripotency induction, and whether

they can induce pluripotency without the aid of transcription factors.

99

3. INDUCING THE PLURIPOTENT STATE

Current and future directions: Non-viral-derived iPSCs26

.

INTRODUCTION

Due to the risk of insertional mutagenesis, viral transduction has been increasingly

replaced by non-viral methods to generate iPSCs. Non-viral methods for generating iPSCs

using adenovirus204

, plasmids205

, or excision of reprogramming factors using Cre/LoxP206, 207

or piggyBAC transposition208

have been reported, but suffer from low reprogramming

efficiencies (< 0.003%) and may leave behind residual vector sequences. Recently, successful

reprogramming of human neonatal foreskin fibroblasts was reported using episomal vectors

derived from the Epstein-Barr virus209

. However, this technique requires three individual

plasmids carrying a total of seven factors, including the oncogene SV40, and has not been

shown to successfully reprogram cells from adult donors, a more clinically-relevant target

population. Further, expression of the EBNA1 protein, as is required for this technique, may

increase immune cell recognition of transfected cells210

, thus potentially limiting clinical

application if the transgene is not completely removed. Protein-based iPSC generation in

mouse211

and human212

fetal and neonatal cells has also been demonstrated, but requires either

chemical treatment (valproic acid)211

or many rounds of treatment212

. Lastly, protein-based

methods require fusion of cell-penetrating peptides to reprogramming factors, a technique

which may be unfamiliar to some laboratories. Most DNA-based methods require only

minimal molecular biology background, and so remain a more attractive option for a wider

population of researchers interested in cellular reprogramming.

Minicircle vectors are supercoiled DNA molecules that lack a bacterial origin of

replication and antibiotic resistance gene; therefore, they are primarily composed of a

eukaryotic expression cassette. Compared to plasmids, minicircle vectors benefit from higher

100

transfection efficiencies and longer ectopic expression due to their lower activation of

exogenous silencing mechanisms213,214

, and thus represent an ideal mechanism for generating

iPSCs.

MATERIALS AND METHODS

We constructed a minicircle plasmid that contained a single cassette of four

reprogramming factors (OCT4, SOX2, LIN28, NANOG), each separated by self-cleaving

peptide 2A sequences26

. For microarray analysis, total RNA was prepared from biological

duplicate samples of hASCs and H7 hESCs, and minicircle-derived iPSC (mc-iPSC)

subclones from two individuals, for a total of six unique samples; samples were then

hybridized and scanned on Agilent Whole Human Genome microarrays. Because the majority

of microarray data found in the Gene Expression Omnibus (GEO) repository

(http://www.ncbi.nlm.nih.gov/geo/) is in the Affymetrix format, we also performed

transcriptional profiling using GeneChip® Human Genome U133 Plus 2.0 Arrays

(Affymetrix), this time using mc-iPSC sublcones from three individuals. Using GeneSpring

GX 10.0, the resulting data were compared to microarray data published by the James

Thomson lab for their non-virally derived human iPSCs209

, and also to our lab‟s data for

lentivirally-derived iPSCs from hASCs27

.

RESULTS AND DISCUSSION

Using the non-viral minicircle reprogramming construct, we successfully derived

transgene-free iPSC colonies (mc-iPSCs) after transfection of hASCs, and performed a

microarray study to compare their transcriptional profiles with that of hESCs and other iPSC

lines derived using lentiviral vectors26

. The results of our global gene expression profiling of

mc-iPSCs, hESCs, and iPSCs derived using both lentiviral and non-viral vectors is shown in

101

Fig. 26. While the transcriptomes of the mc-iPSC subclones cluster closely with those of viral-

derived iPSCs and with hESCs, they are certainly not identical. This heterogeneity of the iPSC

transcriptome, whether viral- or non-viral-derived, distinguishes iPSCs from hESCs for poorly

understood reasons. Similar results as this study have been reported previously for viral-

derived iPSCs215

. This is an interesting finding since iPSCs have all the functional hallmarks

of embryonic cells – self-renewal, pluripotency, etc. – yet they express genes at different

levels and in different patterns. It does not appear however that the core pluripotency genes

(OCT4, SOX2, NANOG) are significantly different24

(see also Fig. 26A). Given the apparent

stochastic nature of current methods for reprogramming, it is perhaps not surprising then that

we see a corresponding heterogeneity in the iPSC transcriptomes. It is not known whether the

primary source of heterogeneity is due to differences in the reprogramming methods used (e.g.

different cocktails of reprogramming factors, viral vs. non-viral) or due to variations in the

starting donor cells themselves. Using next generation DNA and RNA sequencing, our group

intends to find the exact genetic sources that contribute to this heterogeneity in iPSCs.

In the mean time, as the reprogramming field moves from basic science towards

clinical translation, efficient derivation of iPSCs that are free of foreign or chemical elements

is absolutely critical. Here, we describe a simple method for generating transgene-free iPSCs

from adult donor sources, and that requires only a single vector without the need for

subsequent drug selection or vector-excision, or inclusion of oncogenes such as SV40. In the

future, it will be important to ensure the complete absence of minicircle DNA in therapeutic

cells that have been differentiated from mc-iPSCs. Finally, with its basic molecular principles

and straightforward protocol, minicircle DNA is ideally suited for facilitating iPSC research

around the world.

102

Figure 26. Microarray analysis of non-viral-derived iPSCs. (A) Upper panel, heat map

showing two mc-iPSC subclones are similar to H7 human ESCs and distinct from hASCs.

Lower panel, scatter plots depicting the range of gene fold changes between paired cell

types (the iPSC data is the average of both subclones). Highlighted are OCT4, SOX2, and

NANOG expression (red arrows). Green lines indicate five-fold changes in expression

levels between samples. (B) Pearson hierarchical clustering of mc-iPSCs with other

pluripotent and somatic cell types. Averaged microarray data sets for “Foreskin”, “hESC”

(H7, H9, H13, H14 lines), “JT-iPSC” (iPSCs derived from foreskin cells using non-viral

methods209

) were downloaded from the GEO repository. “lenti-iPSC” represents averaged

microarray data for iPSCs derived from hASCs using standard lentiviral methods27

. mc-

iPSC-1s, mc-iPSC-2s, and mc-iPSC-3s represent microarray data for individual subclones

(from three different donors) of minicircle-derived iPSCs.

103

3. INDUCING THE PLURIPOTENT STATE

Current and future directions: Comparison of induced pluripotent stem cells that

have been derived from different tissue sources25

.

INTRODUCTION

Human induced pluripotent stem cells (iPSCs) generated by de-differentiation of adult

somatic cells offer potential solutions for the ethical issues surrounding human embryonic

stem cells (hESCs), as well as their immunologic rejection after cellular transplantation.

However, although human iPSCs have been described as “embryonic stem cell-like”, these

cells have a distinct gene expression pattern compared to hESCs, making incomplete

reprogramming a potential pitfall. It is unclear to what degree the difference in tissue of origin

may contribute to these gene expression differences. In this study, we use microarray data

from our own lab as well as from public sources to compare human iPSCs that have been

derived from different tissue sources. Specifically, we compare the gene profiles of human

iPSCs derived from foreskin fibroblasts209

, neonatal female fibroblasts212

, adipose stem cells27

,

and keratinocytes216

with their corresponding donor cells, and show that there is a

transcriptional “memory” in iPSCs of the donor cell.

MATERIALS AND METHODS

Gene expression data was obtained from the Gene Expression Omnibus (GEO)

repository. In addition to the iPSCs and donor cells mentioned above27, 209, 212, 216

, we included

six unique hESC lines in our analysis (H1, H7, H9, H13, H14, and T3), all of which are also

derived from GEO27, 209, 212, 216

. After normalization of the array data, we defined the distance

metric between two samples to be the percentage of genes that are differentially expressed

between them (P<0.05, Fold change>2.0). Thus, two “closer” groups will have a lower

percentage of genes that are different between them, and vice versa.

104

RESULTS AND DISCUSSION

Based on the calculated distances that represent the overall difference between cell

types, we found that iPSCs are closer to their corresponding donor cell type than to other

donor cell types (Fig. 27, see Ghosh et al.25

for further analysis). This pattern suggests that

there is a residual gene expression of the donor cell that contributes significantly to the

differences among human iPSCs and hESCs, and likely represents incomplete reprogramming.

Furthermore, it appears that some iPSCs are closer to hESCs than others (see “iPS-hFFib” in

Fig. 27A), whereas others are more distant (see “iPS-hKT” in the same figure). The exact

reasons for this are unknown, though it is interesting that the corresponding donor cells

(“hFFib” and “hKT”) are also closest and most distant, respectively, to hESCs (Fig. 27B).

Because the data used in this study comes from multiple laboratories, it is possible

that these transcriptome differences may be attributed to varying culturing conditions and

reprogramming techniques. A second criticism is that there is the potential for contamination

of iPSC populations with incompletely reprogrammed cells, resulting in continued expression

of genes associated with the donor cells217-219

. Irrespective of these criticisms, inherent

differences have been reported between iPSCs and hESCs, and among iPSCs themselves215, 220, 221

.

In summary, while the overall transcriptional profiles of iPSCs from different sources

share a common signature with hESCs, the profiles also suggest retention of memory of the

tissue of origin. Further work to standardize the culture conditions, reprogramming protocols,

and purification of iPSCs is warranted in order to confirm these preliminary findings, and also

to investigate whether persistent donor cell gene expression in iPSCs causes functional

differences in their degree of pluripotency and capacity to differentiate. For now, we must

remain wary of the possibility that the donor cell source may affect the very nature of derived

iPSCs. Results from this study, and from future work, will have profound implications for

clinical regenerative therapies where homogenous, reproducible cell populations are critical.

105

Figure 27. Persistent donor cell gene expression among iPSCs contributes to differences

with hESCs. (A) The distance between hESCs and human iPSCs shows iPS-hFFib to be

closest to hESCs. (B) The distance between hESCs and donor cell types shows hFFib to be

closest to hESCs. The distance between human iPSCs and 4 different donor cell types shows

(C) iPS-hFFib closest to hFFib; (D) iPS-hASC closest to hASC; (E) iPS-hNFib closest to

hNFib; and (F) iPS-hKT closest to hKT. (Closest grouping is marked with a red circle).

(hFFib=human foreskin fibroblast; hASC=human adipose stem cell; hNFib=human female

neonatal fibroblast; hKT=human keratinocyte).

106

ADDENDUM

Transcriptome alteration in the diabetic heart by rosiglitazone: Implications for

cardiovascular mortality28

ABSTRACT

Recently, the type 2 diabetes medication, rosiglitazone, has come under scrutiny for possibly

increasing the risk of cardiac disease and death. To investigate the effects of rosiglitazone on

the diabetic heart, we performed cardiac transcriptional profiling and imaging studies of a

murine model of type 2 diabetes, the C57BL/KLS-leprdb

/leprdb

(db/db) mouse. We compared

cardiac gene expression profiles from three groups: untreated db/db mice, db/db mice after

rosiglitazone treatment, and non-diabetic db/+ mice. Prior to sacrifice, we also performed

cardiac magnetic resonance (CMR) and echocardiography. As expected, overall the db/db

gene expression signature was markedly different from control, but to our surprise was not

significantly reversed with rosiglitazone. In particular, we have uncovered a number of

rosiglitazone modulated genes and pathways that may play a role in the pathophysiology of

the increase in cardiac mortality as seen in several recent meta-analyses. Specifically, the

cumulative upregulation of (1) a matrix metalloproteinase gene that has previously been

implicated in plaque rupture, (2) potassium channel genes involved in membrane potential

maintenance and action potential generation, and (3) sphingolipid and ceramide metabolism-

related genes, together give cause for concern over rosiglitazone‟s safety. Lastly, in vivo

imaging studies revealed minimal differences between rosiglitazone-treated and untreated

db/db mouse hearts, indicating that rosiglitazone‟s effects on gene expression in the heart do

not immediately turn into detectable gross functional changes. This study maps the genomic

expression patterns in the hearts of the db/db murine model of diabetes and illustrates the

impact of rosiglitazone on these patterns. The db/db gene expression signature was markedly

different from control, and was not reversed with rosiglitazone. A smaller number of unique

107

and interesting changes in gene expression were noted with rosiglitazone treatment. Further

study of these genes and molecular pathways will provide important insights into the cardiac

decompensation associated with both diabetes and rosiglitazone treatment.

108

INTRODUCTION

Cardiovascular disease is the leading cause of morbidity and mortality in patients with

type 2 diabetes222

. While atherosclerotic coronary artery disease is highly prevalent in many

diabetic patients, the occurrence of nonischemic cardiomyopathy (“diabetic cardiomyopathy”)

suggests that other processes such as microangiopathy, metabolic factors, or myocardial

fibrosis222

might be involved. Thiazolidinediones (TZDs) such as rosiglitazone are insulin

sensitizers that may also have beneficial properties for diabetic cardiomyopathy. These

agonists bind to a subfamily of nuclear hormone receptors, the peroxisome proliferator-

activated receptors (PPARs, including α, β/δ and γ isoforms, of which the γ isoform is the

most common target), and activate transcription factors that modulate gene expression,

ultimately leading to increased insulin sensitivity in peripheral tissues through poorly defined

pathways223, 224

. However, a recent meta-analysis of clinical trial outcomes for the PPARγ

agonist rosiglitazone found a significant increase in the risk of myocardial infarction, and a

borderline significant risk of death from cardiovascular causes225

. The reasons for this

increased risk are unclear, and a better understanding of the effects of rosiglitazone on the

diabetic heart is urgently needed.

Despite the large number of studies on rosiglitazone and diabetes, the transcriptional

network changes by which PPARγ agonists and other TZDs promote insulin sensitivity are not

well characterized. A number of groups have studied gene expression in the db/db mouse

liver226

, as well as transcriptional changes induced by TZDs in adipocytes, kidney, aorta, and

pancreatic islet cells in various mouse models227-230

, but none has studied the heart directly.

Moreover, determining PPAR-agonist target tissues is complicated by the fact that the three

PPAR isoforms are not distributed equally across all tissues. In adipose tissue, PPARγ

predominates and promotes differentiation and lipid storage231

, resulting in suppression of

lipolysis by insulin and reduction of plasma free fatty acid concentrations. In the

109

cardiovascular system, PPARγ is expressed in vascular smooth muscle and vascular

endothelium232

, but is believed to be in low abundance in cardiomyocytes233

.

Animal models of type 2 diabetes present an optimal system for studying the effects

of rosiglitazone. A well-established murine model is the C57BL/KLS-leprdb

/leprdb

(db/db)

mouse that has a mutation in the leptin receptor. Homozygous db/db mice become obese by 3-

4 wk of age and develop hyperglycemia at 4-8 wk. Serum insulin levels increase as early as

10-14 days, peak at 6-8 wk, then decrease afterward. These mice continue to be

hyperinsulinemic throughout life and ultimately develop cardiomyopathy (contractile

dysfunction), as evidenced from metabolic experiments using cultured db/db

cardiomyocytes234

, echocardiographic studies235

, isolated working heart preparations236

, and

direct in situ ventricular pressure measurements237

. Recent studies by our lab and others have

delineated the longitudinal structural and metabolic cardiomyopathic changes in the db/db

mouse using cardiac magnetic resonance (CMR) and Fluorine-18–2-fluoro-2-deoxy-d-glucose

([18

F]FDG) PET scanning238

.

To address the many questions concerning the molecular and functional effects of

rosiglitazone on the diabetic heart, we used CMR imaging and echocardiagraphic studies to

assess the development of cardiomyopathy in untreated and rosiglitazone-treated db/db mice,

as well as normal db/+ mice. These studies were followed by cardiac gene expression

profiling of the diabetic and control phenotypes to understand the underlying molecular

changes in heart tissue. A comparison of the cardiac transcriptomes of rosiglitazone-treated

db/db mice with untreated db/db and normal db/+ mice thus provides insights into the effects

of rosiglitazone on the heart, either through direct or indirect mechanisms, as well as clues to

its potential toxicity.

110

MATERIALS AND METHODS

Animal and sample tissue preparation. Homozygous db/db mice (Jackson Laboratories, Bar

Harbor, ME) were maintained on a normal chow diet and housed in a room with a 12:12-h

light-dark cycle and an ambient temperature of 22°C. For the treatment group, homozygous

db/db mice were maintained on 5 mg/kg/day (approximately 0.225 mg/mouse/day)

rosiglitazone (GlaxoSmithKline, London, UK) that was mixed with normal chow. Unless

otherwise stated, heterozygous db/+ littermates were used as control animals. All protocols

were approved by the Administrative Panel on Laboratory Animal Care at Stanford University

and were carried out in accordance with the guidelines of the American Association for

Accreditation of Laboratory Animal Care. At the end of the study, animals were euthanized

with a lethal dose of isoflurane. Immediately after death, wet heart weight (HW) and body

weight (BW) were measured. Whole hearts, pancreas, and liver were harvested and preserved

in TRIzol reagent (Invitrogen, San Diego, CA) for subsequent mRNA isolation, and a subset

of hearts were fixed in 10% formalin for histological evaluation.

Insulin tolerance testing. Insulin tolerance testing was performed on mice after a 6-h fast. At

the time of testing, a bolus of human regular insulin (Eli Lilly, Indianapolis, IN; 0.75 IU/kg)

was injected intraperitoneally. In blood derived from a tail nick, glucose levels were then

determined with a FreeStyle blood glucose monitoring system (Abbott Laboratories, Abbott

Park, IL) at baseline and 30, 60, and 120 min after injection.

Insulin Enzyme-Linked ImmunoSorbent Assay (ELISA). After a 6-h fast, roughly 200 µl

of blood was obtained from each animal via retroorbital bleeding. Samples were placed in

500- µl tubes containing EDTA (Becton-Dickinson, Franklin Lakes, NJ) and centrifuged at

4°C and 13,200 rpm for 10 min. Approximately 50-75 µl of supernatant (serum) was then

111

collected for further processing. Serum insulin concentrations were measured with a Mercodia

mouse insulin enzyme immunoassay kit (Alpco Diagnostics, Salem, NH).

Nonesterified fatty acids (NEFA). Plasma nonesterified fatty acid concentrations were

measured using a HR series NEFA-HR(2) colorimetric assay kit (Wako Chemicals USA,

Richmond, VA).

Left ventricular functional analysis with echocardiogram. Echocardiography was

performed by a blinded investigator (ZL) using the Siemens-Acuson Sequioa C512 system

equipped with a multi-frequency (8-14 MHz) 15L8 transducer. Analysis of M-mode images

was performed using the Siemens built-in software. Left ventricular end diastolic diameter

(EDD) and end-systolic diameter (ESD) were measured and used to calculate fractional

shortening (FS) by the following formula: FS= [EDD-ESD]/EDD. LV volume at end diastolic

(EDV) and end-systole (ESV) were calculated by the bullet method as follows: EDV = 0.85 ×

CSA(d) × L(d), ESV = 0.85 × CSA(s) × L(s), where CSA(d) and (s) are endocardial area in

end-diastole and end-systole, respectively, obtained from short-axis view at the level of the

papillary muscles. L(d) and L(s) are the LV length (apex to mid-mitral annulus plane) in end-

diastole and end-systole, respectively, obtained from the parasternal long-axis view. LV

ejection fraction (EF%) was calculated as: EF% = (EDV - ESV) × 100 / EDV.

Cardiac magnetic resonance imaging. To prepare for scanning, induction of anesthesia was

accomplished with 2% isoflurane and 1 l/min oxygen. Respiratory rate was monitored and

used to manually calibrate the maintenance dose of isoflurane at 1.25–1.5%. Platinum needle

ECG leads were inserted subcutaneously in the right and left anterior chest wall. Respiration

was monitored with a pneumatic pillow sensor positioned along the abdomen. Body

112

temperature was maintained at 36–37°C by a flow of heated air thermostatically controlled by

a rectal temperature probe. Heart rate (HR), respiratory rate, and body temperature were

recorded every 4 min during image acquisition. Magnetic resonance images were acquired by

a blinded investigator (PY) with a 4.7-T magnet (Bruker BioSpin, Fremont, CA) controlled by

a Varian Inova Console (Varian, Palo Alto, CA), using a transmit-receive quadrature volume

coil with an inner diameter of 3.5 cm. For particularly obese animals, a larger coil with an

inner diameter of 6 cm was utilized. Image acquisition was gated to the ECG R-wave (Small

Animal Instruments, Stony Brook, NY). Coronal and axial scout images were used to position

a two-dimensional imaging plane along the short axis of the left ventricular (LV) cavity. Gated

gradient echo sequences were then used to acquire sequential short-axis slices spaced 1 mm

apart from apex to base. For each sequence, 12 cine frames encompassing one cardiac cycle

were obtained at each slice level with the following sequence parameters: acquisition time

(TR) = 100–140 ms, echo time (TE) = 2.8–3.5 ms, number of repeats (NEX) = 8, field of view

(FOV) = 30×30 mm, matrix = 128×128, flip angle = 60°. For each short-axis slice, planimetry

measurements of LV myocardial area were conducted off-line by tracing the epicardial and

endocardial borders at end systole and end diastole with MRVision software (MRVision,

Winchester, MA). For these purposes the papillary muscles were considered part of the LV

cavity. Anteroseptal and posterior wall thickness measurements were performed on a

midventricular slice at the level of the papillary muscles at end diastole. LV mass (LVM) was

derived from the sum of the differences between the end-diastolic epicardial and endocardial

areas from apex to base, adjusted for the specific gravity of myocardial tissue (1.055 g/ml).

LV end-diastolic (LVEDV) and end-systolic (LVESV) volumes were calculated as the sum of

the endocardial areas of each slice from the apex to the LV outflow tract at end diastole and

end systole, respectively. LV ejection fraction (LVEF) was calculated as (LVEDV –

LVESV)/LVEDV. Cardiac output (CO) was calculated as (LVEDV – LVESV)/(HR).

113

RNA isolation and quality control. Total RNA was isolated separately from five heart

samples for each condition using TRIzol reagent (Invitrogen, San Diego, CA) according to the

manufacturer‟s instructions for a total of 15 distinct RNA samples. Total RNA was purified

using RNeasy columns (Qiagen, Chatsworth, CA). RNA concentration was measured by

spectrophotometry, and RNA integrity assessed with an Agilent 2100 bioanalyzer (Agilent

Technologies, Santa Clara, CA) with 6000 Nano Chips according to the manufacturer‟s

instructions. RNA was judged as suitable for array hybridization only if samples exhibited

intact bands corresponding to 18S and 28S ribosomal RNA subunits and had a RNA Integrity

Number (RIN) greater than six. Universal Reference RNAs for mouse (Stratagene, La Jolla,

CA) were purchased as microarray reference controls.

Labeling reaction and hybridization. Using Low RNA Input Fluorescent Linear

Amplification Kits (Agilent Technologies, Santa Clara, CA, USA), cDNA was reverse

transcribed from each RNA sample, and cRNA was then transcribed and fluorescently labeled

from each cDNA sample. The fifteen mouse heart tissue cRNA samples were labeled with

Cy5 and the Universal Mouse reference cRNA was labeled with Cy3, yielding a total of 5

biological replicates per condition. The resulting cRNA was purified using an RNeasy kit

(Qiagen, Valencia, CA, USA) followed by quantification of the cRNA by spectroscopy using

an ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). 825ng

Cy3- and Cy5- labeled and amplified cRNA was mixed and fragmented according to the

Agilent technology protocol. cRNA was hybridized to 4×44K whole human genome

microarray slides from Agilent (Part G4112F) according to the manufacturer‟s instructions.

The hybridization was carried in a rotating hybridization chamber in the dark at 65°C for 17 h.

114

Washing, scanning, and feature extraction. Slides were washed with Gene Expression

Wash Buffer 1 and 2 (Agilent Technologies, Santa Clara, CA) followed by Acetonitrile. A

final wash in Stabilization and Drying Solution was performed to prevent Cy-5 degradation by

ozone. The array was scanned using an Agilent G2505B DNA microarray scanner under

extended Dynamic range. The image files were extracted using the Agilent Feature Extraction

software version 9.5.1 applying LOWESS background subtraction and dye-normalization.

Further analyses were performed with BRB ArrayTools Version 3.4 (National Cancer

Institute) and TIGR MeV software (http://www.tm4.org/).

Real-time quantitative PCR for confirmation of microarray results. RNA to cDNA

conversion was performed with a SuperScript First-Strand cDNA synthesis kit (Invitrogen).

The cDNA was then used as a template in a TaqMan real-time PCR assay with the ABI Prism

7700 Sequence Detection System (Applied Biosystems, Foster City, CA). All samples were

run in triplicates. Specialized premade gene expression assay reagents for potassium channel,

subfamily K, member 1 (KCNK1; catalog no. Mm00434624_m1), matrix metalloproteinase 3

(MMP3; Mm00440295_m1), beta-1,4-N-acetyl-galactosaminyl transferase 1 (B4GALNT1;

Mm00484653_m1), carboxyl ester lipase (CEL; Mm00486975_m1), and N-acylsphingosine

amidohydrolase 2 (ASAH2; Mm00479659_m1), purchased from Applied Biosystems, were

used for these experiments. Threshold cycles were placed in the logarithmic portion of the

amplification curve, and each sample was referenced to 18S RNA (part no. 4319413E)

amplification to control for the total amount of RNA. Fold differences between samples

(relative quantification) were calculated with the delta-delta method [S1/S2 = 2-(T1-T2)

], where

S1 and S2 represent samples 1 and 2 and T1 and T2 denote the threshold cycles for S1 and S2.

115

Histological analysis of cardiac tissue. At the time of death, a select group of hearts were

fixed in 10% formalin, cut at the midventricular level, and embedded in paraffin blocks. The

blocks were then sectioned into short-axis slices, which were subsequently stained with

hematoxylin and eoxin according to standard protocols. High-power fields magnified to 40x

from the midportion of the LV free wall were photographed.

Statistical Analysis. For all tests, one-way ANOVA was performed. All P values <0.05 were

considered significant. For microarray analysis, the Significance Analysis of Microarrays

(SAM) algorithm was used to identify genes with statistically significant differences in

expression among the conditions. SAM is a statistically rigorous test that incorporates a FDR

calculation to correct for multiple testing errors. Gene Ontology group overrepresentation

analysis was performed using Fisher's exact test with FDR correction through High

Throughput GoMiner software.

RESULTS

Insulin resistance in rosiglitazone-treated, untreated and control mice. We

analyzed a number of metabolic parameters to confirm insulin resistance in our db/db murine

model of diabetes before proceeding with microarrays. Insulin resistance is strongly associated

with obesity, and one mechanism may be the generation of metabolic messengers such as free

fatty acids by adipose tissue that inhibit insulin action on muscle239, 240

. We measured plasma

non-esterified fatty acid (NEFA) and insulin levels to follow insulin resistance in our three

groups of mice. Fig. 28A shows plasma NEFA levels in the three groups of mice. Compared

to untreated db/db mice, NEFA levels in the rosiglitazone-treated db/db mice decreased

dramatically. Fasting insulin levels (Fig. 28B) and insulin tolerance testing (Fig. 28C) further

confirmed the improved insulin sensitivity with rosiglitazone treatment.

116

Figure 28. Insulin resistance and mean body/heart weights of rosiglitazone-treated,

untreated, and control mice. (A) Plasma NEFA levels at 0, 1, and 3 months in untreated

db/db mice (n = 10), rosiglitazone-treated db/db mice (n = 10), and db/+ mice (n = 10).

NEFA levels were higher in both the treated and untreated db/db groups at baseline when

compared to db/+ control mice. Compared to untreated db/db mice, NEFA levels in the

rosiglitazone-treated db/db group decreased significantly during the first month (960.9 vs.

1501.0 mmol/L, P<0.05), and decreased further by the third month to near db/+ control

levels. (B) Fasting insulin levels in the three groups at 0, 1, and 3 months. At baseline, both

treated and untreated db/db groups had increased levels of insulin compared to db/+ control.

Insulin levels dramatically increased in untreated db/db mice at 1 month, but progressively

decreased in rosiglitazone-treated mice. (C) Insulin tolerance testing at 3 months. Serum

glucose levels after insulin injection were significantly higher in untreated db/db mice when

compared to db/+ mice; the rosiglitazone-treated db/db group had moderately elevated

glucose levels after insulin administration. (D) Mean body weights of the three groups of

mice at 0, 1, and 3 months. Weights of both treated and untreated db/db mice were markedly

higher than db/+ controls at baseline. Furthermore, rosiglitazone-treated mice had generally

higher mean body weights when compared to untreated mice. (E) Mean heart/body weight

ratios at sacrifice (4 months after treatment initiation). No significant difference between

treated and untreated groups (3.27 vs. 2.58, P = 0.13), though there does appear to be a

downward trend in the data. Heart/body weight ratio was significantly higher in the db/+

group compared to both db/db groups. Values are mean±SEM. *P<0.05 vs. age-matched

db/+; #P<0.05 vs. age-matched untreated db/db.

117

Body and heart morphometric analysis. In addition to the obesity associated with

type 2 diabetes, rosiglitazone and other TZDs are widely known to cause further weight gain

due to fluid retention and altered adipogenesis241

. Mean body weights of both treated and

untreated db/db mice were markedly higher than db/+ controls at baseline (Fig. 28D).

Furthermore, rosiglitazone-treated mice had generally higher mean body weights when

compared to untreated mice, as expected. The heart/body weight ratio was significantly higher

in the db/+ group when compared to the treated and untreated db/db groups (Fig. 28E).

However, histological analysis did not reveal obvious differences at the microscopic level

(Fig. 29).

Left ventricular functional

analysis with

echocardiogram. Previous

groups have observed a

significant decrease in left

ventricular ejection fraction

(LVEF) and fractional

shortening (FS) in db/db

mice when compared with

db/+ controls235, 242

. In one

study, the FS was reduced

by as much as 16% at 12

weeks of age235

. However,

more accurate CMR studies

of cardiac contractility have

not confirmed the dramatic

Figure 29. High-power light microscopy slides of

myocardial tissue from rosiglitazone-treated and

untreated db/db mice, and db/+ controls. Individual

cardiomyocytes from all three groups appear qualitatively

similar based on morphology and size, though the untreated

db/db cardiomyocytes exhibit very mild thickening. No

increases in inflammatory cells, cell death, fibrosis, or other

processes were observed in the two db/db groups compared

to db/+ controls.

118

differences seen in the echocardiographic studies, observing only an approximately 2%

reduction in LVEF at 22 weeks of age238

. Furthermore, clinical echocardiographic studies of

type 2 diabetic patients treated with rosiglitazone found no significant changes in LVEF after

52 weeks of treatment243

. To investigate cardiac contractility and verify these previous

findings, we assessed cardiac contractility in the three mouse groups using echocardiography

(Fig. 30). The mean LVEF and FS levels of the untreated db/db group were not significantly

different from those in the db/+ group, though there does appear to be a downward trend in

contractility in the db/db group. Further, we found no significant LVEF or FS differences

between the rosiglitazone-treated db/db group and untreated db/db group, confirming the

results of the clinical study that found no evidence of rosiglitazone-induced changes in cardiac

contractility.

Cardiac magnetic resonance imaging. Using cardiac magnetic resonance (CMR)

scanning, our group has previously shown progressive cardiomyopathic changes in db/db mice

when compared to db/+ controls238

. Left ventricular mass (LVM), interventricular septal

thickness (IVST) and posterior wall thickness (PWT) were significantly increased in db/db

mice, while LVEF was only marginally decreased. For the CMR imaging in this new study

(Fig. 31), we included rosiglitazone-treated db/db mice in addition to the untreated db/db and

db/+ groups, and scanned them at 8 weeks after treatment initiation to look for

cardiomyopathic changes. CMR scans revealed significant, albeit subtle, cardiomyopathic

changes in untreated db/db mice compared to db/+ controls (Fig. 31). However, rosiglitazone

treatment resulted in no significant improvements in cardiac contractility, which confirmed

our echocardiographic studies.

Gene expression analysis. Whole-heart RNA from five mice from each of the three

groups after four months with or without treatment was used for microarray analysis. Overall,

there were significant transcriptional differences between the db/+ and both db/db groups,

119

while differences between rosiglitazone-treated and untreated db/db groups were much less

dramatic. Fig. 32A shows a 3-dimensional scaling plot based on principal component analysis

of the expression data, which graphically demonstrates that rosiglitazone did not restore db/db

transcriptomes to the normal db/+ state. Genes that exhibited significant differential

expression across the three groups were identified using the SAM statistical algorithm

(selected genes are listed in Tables 6, 7).

Interesting genes that were significantly upregulated in the untreated db/db group

compared to control include PPAR (PPARG, the receptor for rosiglitazone), FK506 binding

Figure 30. Left ventricular functional analysis with echocardiogram of treated and

untreated db/db mice, and db/+ controls. The mean left ventricular ejection fraction

(LVEF) and fractional shortening (FS) of the untreated db/db group (72.4 2.3% and 36.2

1.7%, respectively) were not significantly different from the db/+ group (76.9 4.3%, P =

0.11 and 39.7 3.6%, P = 0.11, respectively), though there appears to be a downward trend

in contractility between these two groups. Further, there were no significant LVEF or FS

differences between the rosiglitazone-treated db/db group (74.3 3.9% and 37.8 3.1%,

respectively) and untreated db/db group (P = 0.43 and 0.40, respectively). Values are

mean±SEM.

120

proteins (FKBP5 and FKBP10), several potassium channel proteins (KCNK1 and KCND3),

two matrix metalloproteinases (MMP3 and MMP8), a cyclin-dependent kinase inhibitor

(CDKN1), and myostatin (GDF8) (Table 6). Specific downregulated genes of interest include

the anti-apoptotic gene B-cell lymphoma–leukemia 2 (BCL2), several tumor necrosis factor-

alpha (TNFα) genes and transforming growth factor (TGFβ) genes, vascular endothelial

Figure 31. Cardiac magnetic resonance imaging of treated and untreated db/db mice,

and db/+ controls. CMR imaging at 8 weeks after treatment initiation of rosiglitazone-

treated db/db mice (n = 4), untreated db/db mice (n = 4) and db/+ groups (n = 4). Left

ventricular mass (LVM), interventricular septal thickness (IVST) and posterior wall

thickness (PWT) were significantly increased in both treated and untreated db/db mice

relative to db/+ controls, while LVEF was decreased (though not significantly in the

rosiglitazone-treated group). No significant differences in LVM, IVST, PWT, or LVEF

between rosiglitazone-treated and untreated db/db mice were detected (P = 0.82, 0.66, 0.14,

and 0.88, respectively). Values are mean±SEM. *P<0.05 vs. age-matched db/+.

121

growth factor (VEGFC), adiponectin (ADIPOQ), and apelin (APLN). We confirmed the

microarray results of 4 selected genes with real-time PCR (Fig. 32B,C). When we compared

the db/db rosiglitazone treated group to control, we found that upregulated genes include the

previously-mentioned MMP3 and KCNK1 transcripts, as well as a potassium channel-

Figure 32. Multidimensional scaling plot of the microarray data of hearts from

treated and untreated db/db mice, and db/+ controls. (A) 3-dimensional scaling plot

provides a graphical representation of high-dimensional expression data in low dimensions.

Each point within a “cloud” represents a single microarray, and the similarity within a set of

microarrays is indicated by their physical proximity to one other. As evidenced from the

figure, each group (db/+ control, untreated db/db, and rosiglitazone-treated db/db) clusters

into a distinct grouping, indicating that each group has a similar transcriptome that can be

distinguished from the two other groups. Taqman real-time PCR of four selected genes

normalized to 18s shows (B) significant upregulation of Kcnk1 in untreated db/db vs. db/+

mice and (C) significant upregulation of all four genes in rosiglitazone-treated vs. untreated

db/db mice. These data confirm the altered regulatory patterns of these four genes in the

microarray data. Values are mean±SEM. *P<0.05 vs. control.

122

interacting gene (KCNIP4), ryanodine receptor (Ryr1), Myosin IIIB (MYO3B), Patched

homolog 1, and apelin (Table 7).

While identifying individual genes that are differentially regulated in each of the

conditions is informative, it is also useful to ascertain which cellular processes are up- or

down- regulated; therefore, we performed the Gene Ontology pathway overrepresentation

analysis using Fisher's exact test. Lipid and protein metabolism, fatty acid beta-oxidation, cell

death, apoptosis, peroxisome organization, and biogenesis were significantly upregulated in

untreated db/db mice when compared to control db/+ mice. Among major pathways that were

significantly downregulated are cell proliferation and cycle, immune response, blood vessel

and vasculature development, and anti-apoptosis. Interestingly, comparing the rosiglitazone-

treated db/db mice to control mice; we did not see a reversal of the pathways that were

upregulated in the untreated db/db group. The additional pathways that were significantly

upregulated in the rosiglitazone-treated db/db hearts, however, included secretion, exocytosis,

protein and intracellular transport, cellular protein metabolism, and sphingolipid metabolism.

DISCUSSION

Few studies have studied the global cardiac gene expression changes resulting from

rosiglitazone treatment. Understanding these molecular changes will help elucidate the

potential mechanism(s) for the increased cardiac disease observed in diabetic patients treated

with this drug. Our microarray results demonstrate that rosiglitazone does not reverse many of

the significant gene expression and pathway changes that occur in untreated diabetic hearts,

such as apoptosis and lipid metabolism. Furthermore, our microarray analysis indicates that

rosiglitazone did not upregulate important insulin- However, several studies have shown

PPARγ expression to be important to cardiac metabolism. One study found that

rosiglitazone-treated Zucker fatty rat hearts had improved regulated glucose transporters

123

Table 6. Selected genes that are differentially expressed between the hearts of

untreated db/db mice and db/+ controls.

Gene Symbol RefSeq ID Name Score(d) Fold Change

Kcnk1 NM_008430 Potassium channel, subfamily K, member 1 7.82 26.03

Gdf8 NM_010834 growth differentiation factor 8 (Gdf8) 5.12 15.93

Egfbp2 NM_010115 epidermal growth factor binding protein type B (Egfbp2) 4.77 9.47

Fkbp5 NM_010220 FK506 binding protein 5 9.76 6.11

Mmp8 NM_008611 Matrix metallopeptidase 8 5.22 5.93

Edn3 NM_007903 Endothelin 3 6.59 5.72

Cdkn1a NM_007669 Cyclin-dependent kinase inhibitor 1A (P21) 5.13 3.74

Kcnj4 NM_008427 Potassium inwardly-rectifying channel, subfamily J, member 4 5.06 2.68

Map3k6 NM_016693 Mitogen-activated protein kinase kinase kinase 6 9.39 2.54

Map3k15 BC031147 Mitogen-activated protein kinase kinase kinase 15 5.88 2.10

Fkbp10 NM_010221 FK506 binding protein 10 5.21 2.09

Mmp3 NM_010809 Matrix metallopeptidase 3 4.30 1.90

Kcnd3 NM_019931 Potassium voltage-gated channel, Shal-related family, member 3 5.93 1.77

Ryr3 XM_619795 Ryanodine receptor 3 5.45 1.74

Pparg NM_011146 Peroxisome proliferator activated receptor gamma 5.51 1.42

Tnfrsf22 NM_023680 Tumor necrosis factor receptor superfamily, member 22 4.06 1.40

Aifm2 NM_153779 Apoptosis-inducing factor, mitochondrion-associated 2 5.62 1.39

Il4ra NM_001008700 interleukin 4 receptor, alpha (Il4ra) 4.96 1.37

Sod1 BC057074 superoxide dismutase 1, soluble, partial cds. 4.02 1.34

Adipor2 NM_197985 Adiponectin receptor 2 4.42 1.22

Myo9a AK029836 Myosin IXa -4.70 -1.26

Casp7 NM_007611 Caspase 7 -4.46 -1.27

Casp2 NM_007610 caspase 2 (Casp2) -5.10 -1.34

Vegfc NM_009506 Vascular endothelial growth factor C -6.20 -1.39

Il13ra1 NM_133990 Interleukin 13 receptor, alpha 1 -4.24 -1.40

Gja1 NM_010288 Gap junction membrane channel protein alpha 1 -4.85 -1.49

Il13ra1 NM_133990 Interleukin 13 receptor, alpha 1 -4.23 -1.50

Myo1b NM_010863 Myosin IB -6.11 -1.57

Tnfaip8 NM_134131 Tumor necrosis factor, alpha-induced protein 8 -4.65 -1.71

Adrb2 NM_007420 Adrenergic receptor, beta 2 -4.91 -1.81

Tnfaip8l1 NM_025566 Tumor necrosis factor, alpha-induced protein 8-like 1 -5.52 -1.88

Il16 NM_010551 Interleukin 16 -24.20 -2.42

Myo1d AK037051 Myosin ID -4.19 -2.45

Apln NM_013912 Apelin -4.51 -2.54

Bcl2 NM_009741 B-cell leukemia/lymphoma 2 (Bcl2), transcript variant 1 -7.36 -2.56

Tnfrsf13c NM_028075 Tumor necrosis factor receptor superfamily, member 13c -6.33 -4.54

Adipoq NM_009605 Adiponectin, C1Q and collagen domain containing -5.31 -4.59

124

Table 7. Selected genes that are differentially expressed between rosiglitazone-treated

and untreated db/db mice.

Gene

Symbol RefSeq ID Name Score(d)

Fold

Change

Kcnip4 NM_030265 Kv channel interacting protein 4 4.46 4.76

Ptch1 NM_008957 Patched homolog 1 6.67 3.32

Ryr1 NM_009109 Ryanodine receptor 1, skeletal muscle 5.27 3.17

Kcnk1 NM_008430 Potassium channel, subfamily K, member 1 6.38 2.84

Myo3b AK033795 adult male epididymis cDNA, RIKEN full-length enriched

library, clone:9230110G05 4.46 2.58

Mmp3 NM_010809 Matrix metallopeptidase 3 4.70 2.03

Tnfsf18 NM_183391 Tumor necrosis factor (ligand) superfamily, member 18 3.82 1.86

Apln NM_013912 Apelin 3.52 1.80

Anxa13 NM_027211 Annexin A13 4.04 1.64

St3gal5 NM_011375 ST3 beta-galactoside alpha-2,3-sialyltransferase 5 5.56 1.62

Col22a1 XM_193814 Collagen, type XXII, alpha 1 3.65 1.45

Col4a1 NM_009931 Procollagen, type IV, alpha 1 4.98 1.26

Kras NM_021284 V-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog 4.40 1.23

B4galnt1 BC022180 Beta-1,4-N-acetyl-galactosaminyl transferase 1 7.83 1.44

Cel NM_009885 Carboxyl ester lipase 6.31 1.38

Pparbp NM_013634 Peroxisome proliferator activated receptor binding protein 3.68 1.29

Tnfrsf10b NM_020275 Tumor necrosis factor receptor superfamily, member 10b 3.54 1.27

Asah2 NM_018830 N-acylsphingosine amidohydrolase 2 4.38 1.21

Myh9 NM_022410 Myosin, heavy polypeptide 9, non-muscle 4.03 1.19

Bdkrb1 NM_007539 Bradykinin receptor, beta 1 -6.72 -1.96

Hsf1 NM_008296 Heat shock factor 1 -5.70 -1.17

Col23a1 AK162470

12 days embryo embryonic body between diaphragm region and

neck cDNA, RIKEN full-length enriched library,

clone:9430076L03

-5.70 -1.13

such as GLUT4 in the heart, a finding that has been reported previously in isolated rat

cardiomyocytes244

. However, we observed a number of unique genes which appear to be

differentially regulated between rosiglitazone-treated and untreated animals that may give

mechanistic insights into rosiglitazone‟s effects on the diabetic heart.

We were initially puzzled to find that PPARγ expression was upregulated in the hearts

of untreated diabetic mice, as it has been reported to be low in cardiomyocytes233, 245, 246

.

125

glucose uptake as well as improved contractility compared to lean rat hearts240

. Another study

found that mice lacking PPARγ in cardiomyocytes exhibited mild cardiac hypertrophy247

.

These findings may reflect an indirect effect of rosiligitazone on the heart, as demonstrated by

a study in which metabolic changes in PPARγ-treated db/db hearts were thought to be

secondary to changes in the supply of exogenous substrates242

. It appears that cardiac PPARγ

expression may play an important though poorly understood role in cardiac physiology, and it

is possible that modulation of PPARγ signaling may affect the response to cardiac ischemia.

Matrix metalloproteinases are known to play a critical role in atherosclerosis and

cardiovascular tissue remodeling, mediating the balance between matrix accumulation and

degradation248

. We found that rosiglitazone caused significant upregulation of MMP3, which

encodes the matrix metalloproteinase stromelysin. MMP3 is typically expressed by

macrophages within the atherosclerotic plaques found in coronary artery disease249

, and has

been shown to promote plaque rupture, myocardial infarction and aneurysm250

. It is therefore

possible that the in vivo upregulation of MMP3 by rosiglitazone will promote plaque rupture,

leading to increased rates of myocardial infarction, as well as increasing tissue remodeling

after ischemia.

In both untreated and rosiglitazone-treated db/db mice, we found several potassium

channel-related genes that were highly upregulated, including KCNIP4 and KCNK1. These

findings support the hypothesis that cardiomyocytes from db/db hearts exhibit

electrophysiological alterations such as attenuated outward K+

currents242, 251

. In general,

glycolytic ATP production is important for the normal function of cardiac membrane ion

channels and pumps, and inhibition of glycolysis causes altered intracellular ion

concentrations that may affect cardiac action potentials. Specifically, KCNIP4 modulates A-

type potassium channels and has been shown to increase expression of cardiac A-type Kv-4

channels252

. KCNK1 encodes a two-pore potassium channel, TWIK-1, which plays a major

126

role in setting the resting membrane potential in many cell types253

. Although it is unclear to

what degree cardiac arrhythmias might have played a role in the mortality shown in the meta-

analysis of rosiglitazone trials225

, the changes we observed in cardiomyocyte potassium

channel expression raise some concerns.

Rosiglitazone treatment also upregulated four genes related to sphingolipid

metabolism: B4GALNT1, CEL, ASAH2, and ST3GAL5. Found in eukaryotic plasma

membranes, sphingolipids have emerged as a class of lipid mediators believed to be important

for endogenous modulation of cardiovascular function254, 255

. Generation of sphingolipid

metabolites such as ceramide, sphingosine, and sphingosine-1-phosphate can activate smooth

muscle cell proliferation, endothelial cell differentiation, apoptosis and cell death, migration,

vasoconstriction, and dilation. Specifically, ceramide is an important mediator of lipotoxicity

in the heart, and accumulation of this metabolite has been associated with cardiomyocyte

apoptosis256

. Recent evidence indicates that the TZD pioglitazone causes increased de novo

ceramide synthesis in rat hearts257

, although another group has reported decreased ceramide

accumulation and subsequent reduction in apoptosis258

. Regardless of the specific mechanism

involved, our microarray data suggest a significant role for sphingolipid metabolism in the

rosiglitazone-treated diabetic heart which we are actively investigating further.

The downregulation of the hormone adiponectin in the diabetic state has been well

established in adipocytes and serum259

, but it was not until 2005 that Pineiro et al. showed

evidence that cardiomyocytes also synthesize adiponectin260

. Confirming a recent report by

Natarajan et al.261

, our results demonstrate a reduction of adiponectin in the diabetic heart,

though we did not observe restoration of its expression after rosiglitazone treatment. In

contrast, the signaling peptide apelin, a recently discovered regulator of cardiac and vascular

function262

, was downregulated in the untreated db/db group but was restored to normal levels

after rosiglitazone treatment. Apelin is expressed in the endothelium of heart, kidney, and

127

lung, and its receptor (APJ) is expressed by myocardial cells and some vascular smooth

muscle cells. Apelin has inotropic effects on cardiac contractility and has been shown to

decrease systemic vascular resistance, and is therefore likely beneficial during heart failure. In

fact, apelin signaling may augment inhibition of the renin-angiotensin system263

, which is

known to exacerbate heart failure when left unchecked. Though not well understood, the

altered expression of apelin we observed in the diabetic heart in response to rosiglitazone may

present a novel pathway for the study of diabetic cardiomyopathy.

In conclusion, this study has mapped the genomic expression patterns in the hearts of

the db/db murine model of diabetes and the effects of rosiglitazone on these patterns. Overall,

as we expected, the db/db gene expression signature was markedly different from control, but

to our initial surprise was not significantly reversed with rosiglitazone. In fact, many of the

transcriptional changes induced by diabetic disease on the heart appeared to be further

exacerbated by rosiglitazone. In particular, we have highlighted a number of rosiglitazone

modulated genes and pathways that may play a role in the pathophysiology of the increase in

cardiac mortality seen in the meta-analysis by Nissen et al. The cumulative upregulation of (1)

a matrix metalloproteinase gene that has previously been implicated in plaque rupture, (2)

potassium channel genes involved in membrane potential maintenance and action potential

generation, and (3) sphingolipid and ceramide metabolism related genes, all give cause for

concern over rosiglitazone‟s safety. Interestingly, a second meta-analysis has suggested that

pioglitazone, also a TZD, may not have the same negative cardiovascular effects as

rosiglitazone264

. We believe the overall expression changes caused by rosiglitazone in the

heart will likely also be found with pioglitazone treatment, though with some differences. This

is based on previous microarray studies of adipocytes265

and hepatocytes266

that found similar,

but not identical, expression profiles after treatment with pioglitazone, rosiglitazone, and other

TZDs. Future investigations of how these two drugs differ at the transcriptional level in the

128

heart will further elucidate the networks that govern diabetic cardiovascular function and

disease.

129

LIST OF REFERENCES

1. Thomson JA, Itskovitz-Eldor J, Shapiro SS, Waknitz MA, Swiergiel JJ, Marshall VS,

Jones JM. Embryonic stem cell lines derived from human blastocysts. Science.

1998;282(5391):1145-1147.

2. Swijnenburg R-J, Schrepfer S, Govaert JA, Cao F, Ransohoff K, Sheikh AY, Haddad

M, Connolly AJ, Davis MM, Robbins RC, Wu JC. Immunosuppressive therapy

mitigates immunological rejection of human embryonic stem cell xenografts.

Proceedings of the National Academy of Sciences of the United States of America.

2008;105(35):12991-12996.

3. Takahashi K, Yamanaka S. Induction of pluripotent stem cells from mouse embryonic

and adult fibroblast cultures by defined factors. Cell. 2006;126(4):663-676.

4. Takahashi K, Tanabe K, Ohnuki M, Narita M, Ichisaka T, Tomoda K, Yamanaka S.

Induction of pluripotent stem cells from adult human fibroblasts by defined factors.

Cell. 2007;131(5):861-872.

5. Yu J, Vodyanik MA, Smuga-Otto K, Antosiewicz-Bourget J, Frane JL, Tian S, Nie J,

Jonsdottir GA, Ruotti V, Stewart R, Slukvin II, Thomson JA. Induced pluripotent

stem cell lines derived from human somatic cells. Science. 2007;318(5858):1917-

1920.

6. Park IH, Arora N, Huo H, Maherali N, Ahfeldt T, Shimamura A, Lensch MW, Cowan

C, Hochedlinger K, Daley GQ. Disease-specific induced pluripotent stem cells. Cell.

2008;134(5):877-886.

7. Sanger F, Nicklen S, Coulson AR. DNA sequencing with chain-terminating inhibitors.

Proceedings of the National Academy of Sciences of the United States of America.

1977;74(12):5463-5467.

8. Venter J, Adams M, Myers E, Li P, Mural R, Sutton G, Smith H, Yandell M, Evans C,

Holt R, Gocayne J, Amanatides P, Ballew R, Huson D, Wortman J, et al. The

sequence of the human genome. Science. 2001;291(5507):1304-1351.

9. Lander E, Linton L, Birren B, Nusbaum C, Zody M, Baldwin J, Devon K, Dewar K,

Doyle M, FitzHugh W, Funke R, Gage D, Harris K, Heaford A, Howland J, et al.

Initial sequencing and analysis of the human genome. Nature. 2001;409(6822):860-

921.

10. Schena M, Shalon D, Davis RW, Brown PO. Quantitative monitoring of gene

expression patterns with a complementary DNA microarray. Science.

1995;270(5235):467-470.

11. Morozova O, Hirst M, Marra M. Applications of new sequencing technologies for

transcriptome analysis. Annual Review of Genomics and Human Genetics.

2009;10:135-151.

12. Wang Z, Gerstein M, Snyder M. RNA-Seq: a revolutionary tool for transcriptomics.

Nature Reviews. Genetics. 2009;10(1):57-63.

13. Mortazavi A, Williams B, McCue K, Schaeffer L, Wold B. Mapping and quantifying

mammalian transcriptomes by RNA-Seq. Nature Methods. 2008;5(7):621-628.

14. Cloonan N, Forrest ARR, Kolle G, Gardiner BBA, Faulkner G, Brown M, Taylor D,

Steptoe A, Wani S, Bethel G, Robertson A, Perkins A, Bruce S, Lee C, Ranade S, et

al. Stem cell transcriptome profiling via massive-scale mRNA sequencing. Nature

Methods. 2008;5(7):613-619.

15. Wu J, Habegger L, Noisa P, Szekely A, Qiu C, Hutchison S, Raha D, Egholm M, Lin

H, Weissman S, Cui W, Gerstein M, Snyder M. Dynamic transcriptomes during

130

neural differentiation of human embryonic stem cells revealed by short, long, and

paired-end sequencing. Proceedings of the National Academy of Sciences of the

United States of America.107(11):5254-5259.

16. Morin RD, O'Connor MD, Griffith M, Kuchenbauer F, Delaney A, Prabhu A-L, Zhao

Y, McDonald H, Zeng T, Hirst M, Eaves CJ, Marra MA. Application of massively

parallel sequencing to microRNA profiling and discovery in human embryonic stem

cells. Genome Research. 2008;18(4):610-621.

17. Bar M, Wyman SK, Fritz BR, Qi J, Garg KS, Parkin RK, Kroh EM, Bendoraite A,

Mitchell PS, Nelson AM, Ruzzo WL, Ware C, Radich JP, Gentleman R, Ruohola-

Baker H, et al. MicroRNA discovery and profiling in human embryonic stem cells by

deep sequencing of small RNA libraries. Stem Cells. 2008;26(10):2496-2505.

18. Wilson KD, Sun N, Huang M, Zhang WY, Lee AS, Li Z, Wang SX, Wu JC. Effects

of ionizing radiation on self renewal and pluripotency of human embryonic stem cells.

Cancer Research. 2010;(in press).

19. Chen MQ, Xie X, Wilson KD, Sun N, Wu JC, Giovangrandi L, Kovacs GTA.

Current-controlled electrical point-source stimulation of embryonic stem cells.

Cellular and Molecular Bioengineering. 2009;2(4):625-635.

20. Cao F, Wagner RA, Wilson KD, Xie X, Fu J-D, Drukker M, Lee A, Li RA, Gambhir

SS, Weissman IL, Robbins RC, Wu JC. Transcriptional and functional profiling of

human embryonic stem cell-derived cardiomyocytes. PloS One. 2008;3(10):e3474.

21. Wilson KD, Hu S, Venkatasubrahmanyam S, Fu J-D, Sun N, Abilez O, Baugh JJA,

Jia F, Ghosh Z, Li RA, Butte AJ, Wu JC. Dynamic microRNA expression programs

during cardiac differentiation of human embryonic stem cells. Circulation:

Cardiovascular Genetics. 2010;(provisionally accepted).

22. Chun H, Wilson KD, Huang M, Wu JC. Integration of genomics, proteomics, and

imaging for cardiac stem cell therapy. European Journal of Nuclear Medicine and

Molecular Imaging. 2007;34 Suppl 1:S20-S26.

23. Li Z, Wilson KD, Smith B, Kraft D, Jia F, Huang M, Xie X, Robbins R, Gambhir S,

Weissman I, Wu JC. Functional and transcriptional characterization of human

embryonic stem cell-derived endothelial cells for treatment of myocardial infarction.

PloS One. 2009;4(12):e8443-e8443.

24. Wilson KD, Venkatasubrahmanyam S, Jia F, Sun N, Butte AJ, Wu JC. MicroRNA

profiling of human-induced pluripotent stem cells. Stem Cells and Development.

2009;18(5):749-758.

25. Ghosh Z, Wilson KD, Wu Y, Hu S, Quertermous T, Wu JC. Persistent donor cell gene

expression among human induced pluripotent stem cells contributes to differences

with human embryonic stem cells. PloS One. 2010;5(2):e8975-e8975.

26. Jia F, Wilson KD, Sun N, Gupta D, Huang M, Li Z, Panetta N, Chen Z, Robbins RC,

Kay MA, Longaker MT, Wu JC. A nonviral minicircle vector for deriving human iPS

cells. Nature Methods. 2010;7(3):197-199.

27. Sun N, Panetta N, Gupta D, Wilson KD, Lee A, Jia F, Hu S, Cherry A, Robbins RC,

Longaker MT, Wu JC. Feeder-free derivation of induced pluripotent stem cells from

adult human adipose stem cells. Proceedings of the National Academy of Sciences of

the United States of America. 2009;106(37):15720-15725.

28. Wilson KD, Li Z, Wagner RA, Yue P, Tsao P, Nestorova G, Huang M, Hirschberg D,

Yock P, Quertermous T, Wu JC. Transcriptome alteration in the diabetic heart by

rosiglitazone: implications for cardiovascular mortality. PloS One. 2008;3(7):e2609-

e2609.

131

29. Wilson KD, Huang M, Wu JC. Bioluminescence reporter gene imaging of human

embryonic stem cell survival, proliferation, and fate. Methods in Molecular Biology.

2009;574:87-103.

30. Wilson KD, Yu J, Lee A, Wu JC. In vitro and in vivo bioluminescence reporter gene

imaging of human embryonic stem cells. Journal of Visualized Experiments.

2008(14).

31. Yu J, Huang N, Wilson KD, Velotta J, Huang M, Li Z, Lee A, Robbins RC, Cooke J,

Wu JC. nAChRs mediate human embryonic stem cell-derived endothelial cells:

proliferation, apoptosis, and angiogenesis. PloS One. 2009;4(9):e7040-e7040.

32. Brent RL. Utilization of developmental basic science principles in the evaluation of

reproductive risks from pre- and postconception environmental radiation exposures.

Teratology. 1999;59(4):182-204.

33. McCollough CH, Schueler BA, Atwell TD, Braun NN, Regner DM, Brown DL,

LeRoy AJ. Radiation exposure and pregnancy: when should we be concerned?

Radiographics. 2007;27(4):909-917; discussion 917.

34. De Santis M, Cesari E, Nobili E, Straface G, Cavaliere AF, Caruso A. Radiation

effects on development. Birth Defects Research. Part C, Embryo Today.

2007;81(3):177-182.

35. Einstein AJ, Moser KW, Thompson RC, Cerqueira MD, Hanzlova MJ. Radiation dose

to patients from cardiac diagnostic imaging. Circulation. 2007;11(116):1290-1305.

36. Fazel R, Krumholz HM, Wang Y, Ross JS, Chen J, Ting HH, Shah ND, Nasir K,

Einstein AJ, Nallamothu BK. Exposure to low-dose ionizing radiation from medical

imaging procedures. The New England Journal of Medicine. 2009;361(9):849-857.

37. Smith-Bindman R, Lipson J, Marcus R, Kim K-P, Mahesh M, Gould R, de Gonzlez

A, Miglioretti D. Radiation dose associated with common computed tomography

examinations and the associated lifetime attributable risk of cancer. Archives of

Internal Medicine. 2009;169(22):2078-2086.

38. Yaghoubi S, Jensen M, Satyamurthy N, Budhiraja S, Paik D, Czernin J, Gambhir S.

Noninvasive detection of therapeutic cytolytic T cells with 18F-FHBG PET in a

patient with glioma. Nature Clinical Practice Oncology. 2009;6(1):53-58.

39. Narsinh K, Cao F, Wu J. Molecular imaging of human embryonic stem cells. Methods

in Molecular Biology. 2009;515:13-32.

40. Hogan D, Kellison T. Nuclear terrorism. The American Journal of the Medical

Sciences. 2002;323(6):341-349.

41. Aladjem MI, Spike BT, Rodewald LW, Hope TJ, Klemm M, Jaenisch R, Wahl GM.

ES cells do not activate p53-dependent stress responses and undergo p53-independent

apoptosis in response to DNA damage. Current Biology. 1998;8(3):145-155.

42. Hong Y, Stambrook P. Restoration of an absent G1 arrest and protection from

apoptosis in embryonic stem cells after ionizing radiation. Proceedings of the

National Academy of Sciences of the United States of America. 2004;101(40):14443-

14448.

43. Solozobova V, Rolletschek A, Blattner C. Nuclear accumulation and activation of p53

in embryonic stem cells after DNA damage. BMC Cell Biology. 2009;10:46-46.

44. Filion T, Qiao M, Ghule P, Mandeville M, van Wijnen A, Stein J, Lian J, Altieri D,

Stein G. Survival responses of human embryonic stem cells to DNA damage. Journal

of Cellular Physiology. 2009;220(3):586-592.

45. Momcilovic O, Choi S, Varum S, Bakkenist C, Schatten G, Navara C. Ionizing

radiation induces ataxia telangiectasia mutated-dependent checkpoint signaling and

G(2) but not G(1) cell cycle arrest in pluripotent human embryonic stem cells. Stem

Cells. 2009;27(8):1822-1835.

132

46. Qin H, Yu T, Qing T, Liu Y, Zhao Y, Cai J, Li J, Song Z, Qu X, Zhou P, Wu J, Ding

M, Deng H. Regulation of apoptosis and differentiation by p53 in human embryonic

stem cells. The Journal of Biological Chemistry. 2007;282(8):5842-5852.

47. Becker K, Stein J, Lian J, van Wijnen A, Stein G. Establishment of histone gene

regulation and cell cycle checkpoint control in human embryonic stem cells. Journal

of Cellular Physiology. 2007;210(2):517-526.

48. Sokolov M, Panyutin I, Onyshchenko M, Panyutin I, Neumann R. Expression of

pluripotency-associated genes in the surviving fraction of cultured human embryonic

stem cells is not significantly affected by ionizing radiation. Gene. 2010;455(1-2):8-

15.

49. Ludwig TE, Levenstein ME, Jones JM, Berggren WT, Mitchen ER, Frane JL,

Crandall LJ, Daigh CA, Conard KR, Piekarczyk MS, Llanas RA, Thomson JA.

Derivation of human embryonic stem cells in defined conditions. Nature

Biotechnology. 2006;24(2):185-187.

50. Subramanian A, Tamayo P, Mootha V, Mukherjee S, Ebert B, Gillette M, Paulovich

A, Pomeroy S, Golub T, Lander E, Mesirov J. Gene set enrichment analysis: a

knowledge-based approach for interpreting genome-wide expression profiles.

Proceedings of the National Academy of Sciences of the United States of America.

2005;102(43):15545-15550.

51. Shu K-X, Li B, Wu L-X. The p53 network: p53 and its downstream genes. Colloids

and Surfaces B: Biointerfaces. 2007;55(1):10-18.

52. Kobayashi T, Mizuno H, Imayoshi I, Furusawa C, Shirahige K, Kageyama R. The

cyclic gene Hes1 contributes to diverse differentiation responses of embryonic stem

cells. Genes & Development. 2009;23(16):1870-1875.

53. Brachat A, Pierrat B, Xynos A, Brecht K, Simonen M, Brngger A, Heim J. A

microarray-based, integrated approach to identify novel regulators of cancer drug

response and apoptosis. Oncogene. 2002;21(54):8361-8371.

54. Chuang Y-YE, Chen Y, Gadisetti, Chandramouli VR, Cook JA, Coffin D, Tsai M-H,

DeGraff W, Yan H, Zhao S, Russo A, Liu ET, Mitchell JB. Gene expression after

treatment with hydrogen peroxide, menadione, or t-butyl hydroperoxide in breast

cancer cells. Cancer Research. 2002;62:6246-6254.

55. Islaih M, Li B, Kadura I, Reid-Hubbard J, Deahl JT, Altizer J, Watson D, Newton R.

Comparison of gene expression changes induced in mouse and human cells treated

with direct-acting mutagens. Environmental and Molecular Mutagenesis.

2004;44(5):401-419.

56. Besson A, Dowdy S, Roberts J. CDK inhibitors: cell cycle regulators and beyond.

Developmental Cell. 2008;14(2):159-169.

57. Nusse R, Fuerer C, Ching W, Harnish K, Logan C, Zeng A, Berge DT, Kalani Y. Wnt

signaling and stem cell control. Cold Spring Harbor Symposia on Quantitative

Biology. 2008;73:59-66.

58. Russell WL. Mutation frequencies in female mice and the estimation of genetic

hazards of radiation in women. Proceedings of the National Academy of Sciences of

the United States of America. 1977;74(8):3523-3527.

59. Jacquet P. Sensitivity of germ cells and embryos to ionizing radiation. Journal of

Biological Regulators and Homeostatic Agents. 2004;18(2):106-114.

60. Brent RL, Bolden BT. The indirect effect of irradiation on embryonic development. 3.

The contribution of ovarian irradiation, uterine irradiation, oviduct irradiation, and

zygote irradiation to fetal mortality and growth retardation in the rat. Radiation

Research. 1967;30(4):759-773.

133

61. Jacquet P, de Saint-Georges L, Vankerkom J, Baugnet-Mahieu L. Embryonic death,

dwarfism and fetal malformations after irradiation of embryos at the zygote stage:

studies on two mouse strains. Mutation Research. 1995;332(1-2):73-87.

62. Wakeford R, Little MP. Risk coefficients for childhood cancer after intrauterine

irradiation: a review. International Journal of Radiation Biology. 2003;79(5):293-309.

63. Whittington RH, Giovangrandi L, Kovacs GTA. A closed-loop electrical stimulation

system for cardiac cell cultures. IEEE Transactions on Bio-Medical Engineering.

2005;52(7):1261-1270.

64. Hanna LA, Foreman RK, Tarasenko IA, Kessler DS, Labosky PA. Requirement for

Foxd3 in maintaining pluripotent cells of the early mouse embryo. Genes &

Development. 2002;16(20):2650-2561.

65. Loh Y-H, Wu Q, Chew J-L, Vega VB, Zhang W, Chen X, Bourque G, George J,

Leong B, Liu J, Wong K-Y, Sung KW, Lee CWH, Zhao X-D, Chiu K-P, et al. The

Oct4 and Nanog transcription network regulates pluripotency in mouse embryonic

stem cells. Nature Genetics. 2006;38(4):431-440.

66. Okuda A, Fukushima A, Nishimoto M, Orimo A, Yamagishi T, Nabeshima Y, Kuro-o

M, Nabeshima Yi, Boon K, Keaveney M, Stunnenberg HG, Muramatsu M. UTF1, a

novel transcriptional coactivator expressed in pluripotent embryonic stem cells and

extra-embryonic cells. The EMBO Journal. 1998;17(7):2019-2032.

67. Nishimoto M, Miyagi S, Yamagishi T, Sakaguchi T, Niwa H, Muramatsu M, Okuda

A. Oct-3/4 maintains the proliferative embryonic stem cell state via specific binding to

a variant octamer sequence in the regulatory region of the UTF1 locus. Molecular and

Cellular Biology. 2005;25(12):5084-5094.

68. van den Boom V, Kooistra SM, Boesjes M, Geverts B, Houtsmuller AB, Monzen K,

Komuro I, Essers J, Drenth-Diephuis LJ, Eggen BJL. UTF1 is a chromatin-associated

protein involved in ES cell differentiation. The Journal of Cell Biology.

2007;178(6):913-924.

69. Yang DH, Moss EG. Temporally regulated expression of Lin-28 in diverse tissues of

the developing mouse. Gene Expression Patterns. 2003;3(6):719-726.

70. Yang B, Lin H, Xiao J, Lu Y, Luo X, Li B, Zhang Y, Xu C, Bai Y, Wang H, Chen G,

Wang Z. The muscle-specific microRNA miR-1 regulates cardiac arrhythmogenic

potential by targeting GJA1 and KCNJ2. Nature Medicine. 2007;13(4):486-491.

71. Gadue P, Huber TL, Paddison PJ, Keller GM. Wnt and TGF-beta signaling are

required for the induction of an in vitro model of primitive streak formation using

embryonic stem cells. Proceedings of the National Academy of Sciences of the United

States of America. 2006;103(45):16806-16811.

72. de la Cruz CC, Der-Avakian A, Spyropoulos DD, Tieu DD, Carpenter EM. Targeted

disruption of Hoxd9 and Hoxd10 alters locomotor behavior, vertebral identity, and

peripheral nervous system development. Developmental Biology. 1999;216(2):595-

610.

73. Wellik DM, Capecchi MR. Hox10 and Hox11 genes are required to globally pattern

the mammalian skeleton. Science. 2003;301(5631):363-367.

74. Airik R, Bussen M, Singh MK, Petry M, Kispert A. Tbx18 regulates the development

of the ureteral mesenchyme. The Journal of Clinical Investigation. 2006;116(3):663-

674.

75. Bussen M, Petry M, Schuster-Gossler K, Leitges M, Gossler A, Kispert A. The T-box

transcription factor Tbx18 maintains the separation of anterior and posterior somite

compartments. Genes & Development. 2004;18(10):1209-1221.

134

76. Mic FA, Duester G. Patterning of forelimb bud myogenic precursor cells requires

retinoic acid signaling initiated by Raldh2. Developmental Biology. 2003;264(1):191-

201.

77. Kume T, Jiang H, Topczewska JM, Hogan BL. The murine winged helix transcription

factors, Foxc1 and Foxc2, are both required for cardiovascular development and

somitogenesis. Genes & Development. 2001;15(18):2470-2482.

78. Seo S, Kume T. Forkhead transcription factors, Foxc1 and Foxc2, are required for the

morphogenesis of the cardiac outflow tract. Developmental Biology. 2006;296(2):421-

436.

79. Cai C-L, Liang X, Shi Y, Chu P-H, Pfaff SL, Chen J, Evans S. Isl1 identifies a cardiac

progenitor population that proliferates prior to differentiation and contributes a

majority of cells to the heart. Developmental Cell. 2003;5(6):877-889.

80. Laugwitz K-L, Moretti A, Caron L, Nakano A, Chien KR. Islet1 cardiovascular

progenitors: a single source for heart lineages? Development. 2008;135(2):193-205.

81. Karlsson O, Thor S, Norberg T, Ohlsson H, Edlund T. Insulin gene enhancer binding

protein Isl-1 is a member of a novel class of proteins containing both a homeo- and a

Cys-His domain. Nature. 1990;344(6269):879-882.

82. Olson EN. Gene regulatory networks in the evolution and development of the heart.

Science. 2006;313(5795):1922-1927.

83. Yang L, Soonpaa MH, Adler ED, Roepke TK, Kattman SJ, Kennedy M, Henckaerts

E, Bonham K, Abbott GW, Linden RM, Field LJ, Keller GM. Human cardiovascular

progenitor cells develop from a KDR+ embryonic-stem-cell-derived population.

Nature. 2008;453(7194):524-528.

84. Toyofuku T, Zhang H, Kumanogoh A, Takegahara N, Yabuki M, Harada K, Hori M,

Kikutani H. Guidance of myocardial patterning in cardiac development by Sema6D

reverse signalling. Nature Cell Biology. 2004;6(12):1204-1211.

85. Yoshida Y, Han B, Mendelsohn M, Jessell TM. PlexinA1 signaling directs the

segregation of proprioceptive sensory axons in the developing spinal cord. Neuron.

2006;52(5):775-788.

86. Leeman MF, Curran S, Murray GI. The structure, regulation, and function of human

matrix metalloproteinase-13. Critical Reviews in Biochemistry & Molecular Biology.

2002;37(3):149-166.

87. Johansson N, Saarialho-Kere U, Airola K, Herva R, Nissinen L, Westermarck J,

Vuorio E, Heino J, Kahari VM. Collagenase-3 (MMP-13) is expressed by

hypertrophic chondrocytes, periosteal cells, and osteoblasts during human fetal bone

development. Developmental Dynamics. 1997;208(3):387-397.

88. Bi W, Deng JM, Zhang Z, Behringer RR, de Crombrugghe B. Sox9 is required for

cartilage formation. Nature Genetics. 1999;22(1):85-89.

89. Smith N, Dong Y, Lian JB, Pratap J, Kingsley PD, van Wijnen AJ, Stein JL, Schwarz

EM, O'Keefe RJ, Stein GS, Drissi MH. Overlapping expression of Runx1(Cbfa2) and

Runx2(Cbfa1) transcription factors supports cooperative induction of skeletal

development. Journal of Cellular Physiology. 2005;203(1):133-143.

90. Young DW, Pratap J, Javed A, Weiner B, Ohkawa Y, van Wijnen A, Montecino M,

Stein GS, Stein JL, Imbalzano AN, Lian JB. SWI/SNF chromatin remodeling complex

is obligatory for BMP2-induced, Runx2-dependent skeletal gene expression that

controls osteoblast differentiation. Journal of Cellular Biochemistry. 2005;94(4):720-

730.

91. Chen CM, Kraut N, Groudine M, Weintraub H. I-mf, a novel myogenic repressor,

interacts with members of the MyoD family. Cell. 1996;86(5):731-741.

135

92. Lanctot C, Moreau A, Chamberland M, Tremblay ML, Drouin J. Hindlimb patterning

and mandible development require the Ptx1 gene. Development. 1999;126(9):1805-

1810.

93. Eiraku M, Tohgo A, Ono K, Kaneko M, Fujishima K, Hirano T, Kengaku M. DNER

acts as a neuron-specific Notch ligand during Bergmann glial development. Nature

Neuroscience. 2005;8(7):873-880.

94. Cortes-Canteli M, Pignatelli M, Santos A, Perez-Castillo A. CCAAT/enhancer-

binding protein beta plays a regulatory role in differentiation and apoptosis of

neuroblastoma cells. The Journal of Biological Chemistry. 2002;277(7):5460-5467.

95. Rochford JJ, Semple RK, Laudes M, Boyle KB, Christodoulides C, Mulligan C,

Lelliott CJ, Schinner S, Hadaschik D, Mahadevan M, Sethi JK, Vidal-Puig A,

O'Rahilly S. ETO/MTG8 is an inhibitor of C/EBPbeta activity and a regulator of early

adipogenesis. Molecular and Cellular Biology. 2004;24(22):9863-9872.

96. Petryk A, Anderson RM, Jarcho MP, Leaf I, Carlson CS, Klingensmith J, Shawlot W,

O'Connor MB. The mammalian twisted gastrulation gene functions in foregut and

craniofacial development. Developmental Biology. 2004;267(2):374-386.

97. Herrera E, Marcus R, Li S, Williams SE, Erskine L, Lai E, Mason C. Foxd1 is

required for proper formation of the optic chiasm. Development. 2004;131(22):5727-

5739.

98. Hatini V, Huh SO, Herzlinger D, Soares VC, Lai E. Essential role of stromal

mesenchyme in kidney morphogenesis revealed by targeted disruption of Winged

Helix transcription factor BF-2. Genes & Development. 1996;10(12):1467-1478.

99. Wollert KC, Drexler H. Clinical applications of stem cells for the heart. Circulation

Research. 2005;96:151-163.

100. Laflamme MA, Chen KY, Naumova AV, Muskheli V, Fugate JA, Dupras SK,

Reinecke H, Xu C, Hassanipour M, Police S, O'Sullivan C, Collins L, Chen Y,

Minami E, Gill EA, et al. Cardiomyocytes derived from human embryonic stem cells

in pro-survival factors enhance function of infarcted rat hearts. Nature Biotechnology.

2007;25(9):1015-1024.

101. Yang S, Graham J, Kahn JW, Schwartz EA, Gerritsen ME. Functional roles for

PECAM-1 (CD31) and VE-Cadherin (CD144) in tube assembly and lumen formation

in three-dimensional collagen gels. American Journal of Pathology. 1999;155:887-

895.

102. Pan G, Thomson JA. Nanog and transcriptional networks in embryonic stem cell

pluripotency. Cell Research. 2007;17(1):42-49.

103. Chambers I, Colby D, Robertson M, Nichols J, Lee S, Tweedie S, Smith A.

Functional expression cloning of Nanog, a pluripotency sustaining factor in

embryonic stem cells. Cell. 2003;113(5):643-655.

104. Schulz WA, Hoffmann MJ. Transcription factor networks in embryonic stem cells and

testicular cancer and the definition of epigenetics. Epigenetics. 2007;2(1):37-42.

105. Hentschke M, Kurth I, Borgmeyer U, Hubner CA. Germ cell nuclear factor is a

repressor of CRIPTO-1 and CRIPTO-3. Journal of Biological Chemistry.

2006;281(44):33497-33504.

106. Masaki H, Nishida T, Kitajima S, Asahina K, Teraoka H. Developmental

pluripotency-associated 4 (DPPA4) localized in active chromatin inhibits mouse

embryonic stem cell differentiation into a primitive ectoderm lineage. The Journal of

Biological Chemistry. 2007;282(45):33034-33042.

107. Webb GC, Thomas PQ, Ford JH, Rathjen PD. Hesx1, a homeobox gene expressed by

murine embryonic stem cells, maps to mouse chromosome 14, bands A3-B.

Genomics. 1993;18(2):464-466.

136

108. Tian H, McKnight SL, Russell DW. Endothelial PAS domain protein 1 (EPAS1), a

transcription factor selectively expressed in endothelial cells. Genes & Development.

1997;11(1):72-82.

109. Sato TN, Qin Y, Kozak CA, Audus KL. Tie-1 and tie-2 define another class of

putative receptor tyrosine kinase genes expressed in early embryonic vascular system.

Proceedings of the National Academy of Sciences of the United States of America.

1993;90(20):9355-9358.

110. Gale NW, Yancopoulos GD. Growth factors acting via endothelial cell-specific

receptor tyrosine kinases: VEGFs, angiopoietins, and ephrins in vascular

development. Genes & Development. 1999;13(9):1055-1066.

111. Korff T, Braun J, Pfaff D, Augustin HG, Hecker M. Role of ephrinB2 expression in

endothelial cells during arteriogenesis: impact on smooth muscle cell migration and

monocyte recruitment. Blood. 2008;112(1):73-81.

112. Aasheim H-C, Delabie J, Finne EF. Ephrin-A1 binding to CD4+ T lymphocytes

stimulates migration and induces tyrosine phosphorylation of PYK2. Blood.

2005;105(7):2869-2876.

113. Bae J-S, Yang L, Manithody C, Rezaie AR. The ligand occupancy of endothelial

protein C receptor switches the protease-activated receptor 1-dependent signaling

specificity of thrombin from a permeability-enhancing to a barrier-protective response

in endothelial cells. Blood. 2007;110(12):3909-3916.

114. Taddei A, Giampietro C, Conti A, Orsenigo F, Breviario F, Pirazzoli V, Potente M,

Daly C, Dimmeler S, Dejana E. Endothelial adherens junctions control tight junctions

by VE-cadherin-mediated upregulation of claudin-5. Nature Cell Biology.

2008;10(8):923-934.

115. Griffin CT, Srinivasan Y, Zheng YW, Huang W, Coughlin SR. A role for thrombin

receptor signaling in endothelial cells during embryonic development. Science.

2001;293(5535):1666-1670.

116. Matushansky I, Maki RG, Cordon-Cardo C. A context dependent role for Wnt

signaling in tumorigenesis and stem cells. Cell Cycle. 2008;7(6):720-724.

117. Wang H, Charles PC, Wu Y, Ren R, Pi X, Moser M, Barshishat-Kupper M, Rubin JS,

Perou C, Bautch V, Patterson C. Gene expression profile signatures indicate a role for

Wnt signaling in endothelial commitment from embryonic stem cells. Circulation

Research. 2006;98(10):1331-1339.

118. Kaestner KH. The making of the liver: developmental competence in foregut

endoderm and induction of the hepatogenic program. Cell Cycle. 2005;4(9):1146-

1148.

119. Lee CS, Friedman JR, Fulmer JT, Kaestner KH. The initiation of liver development is

dependent on Foxa transcription factors. Nature. 2005;435(7044):944-947.

120. Schroeder TM, Jensen ED, Westendorf JJ. Runx2: a master organizer of gene

transcription in developing and maturing osteoblasts. Birth Defects Research. Part C,

Embryo Today. 2005;75(3):213-225.

121. Molkentin JD, Lin Q, Duncan SA, Olson EN. Requirement of the transcription factor

GATA4 for heart tube formation and ventral morphogenesis. Genes & Development.

1997;11(8):1061-1072.

122. Reiter JF, Alexander J, Rodaway A, Yelon D, Patient R, Holder N, Stainier DY.

Gata5 is required for the development of the heart and endoderm in zebrafish. Genes

& Development. 1999;13(22):2983-2995.

123. Xin M, Davis CA, Molkentin JD, Lien C-L, Duncan SA, Richardson JA, Olson EN. A

threshold of GATA4 and GATA6 expression is required for cardiovascular

137

development. Proceedings of the National Academy of Sciences of the United States of

America. 2006;103(30):11189-11194.

124. Heinonen O, Salonen T, Jalanko A, Peltonen L, Copp A. CLN-1 and CLN-5, genes

for infantile and variant late infantile neuronal ceroid lipofuscinoses, are expressed in

the embryonic human brain. Journal of Comparative Neurology. 2000;426(3):406-

412.

125. Pearson JC, Lemons D, McGinnis W. Modulating Hox gene functions during animal

body patterning. Nature Reviews Genetics. 2005;6(12):893-904.

126. King M, Arnold JS, Shanske A, Morrow BE. T-genes and limb bud development.

American Journal of Medical Genetics. Part A. 2006;140(13):1407-1413.

127. Plageman TF, Yutzey KE. T-box genes and heart development: putting the "T" in

heart. Developmental Dynamics. 2005;232(1):11-20.

128. Marcil A, Dumontier E, Chamberland M, Camper SA, Drouin J. Pitx1 and Pitx2 are

required for development of hindlimb buds. Development. 2003;130(1):45-55.

129. Uyttendaele H, Marazzi G, Wu G, Yan Q, Sassoon D, Kitajewski J. Notch4/int-3, a

mammary proto-oncogene, is an endothelial cell-specific mammalian Notch gene.

Development. 1996;122(7):2251-2259.

130. Uyttendaele H, Ho J, Rossant J, Kitajewski J. Vascular patterning defects associated

with expression of activated Notch4 in embryonic endothelium. Proceedings of the

National Academy of Sciences of the United States of America. 2001;98(10):5643-

5648.

131. Atkins GB, Jain MK. Role of Kruppel-like transcription factors in endothelial biology.

Circulation Research. 2007;100(12):1686-1695.

132. Minshall RD, Sessa WC, Stan RV, Anderson RGW, Malik AB. Caveolin regulation

of endothelial function. American Journal of Physiology. Lung Cellular and

Molecular Physiology. 2003;285(6):L1179-1183.

133. Cowan PJ, Tsang D, Pedic CM, Abbott LR, Shinkel TA, d'Apice AJ, Pearse MJ. The

human ICAM-2 promoter is endothelial cell-specific in vitro and in vivo and contains

critical Sp1 and GATA binding sites. The Journal of Biological Chemistry.

1998;273(19):11737-11744.

134. Jones CA, London NR, Chen H, Park KW, Sauvaget D, Stockton RA, Wythe JD, Suh

W, Larrieu-Lahargue F, Mukouyama Y-S, Lindblom P, Seth P, Frias A, Nishiya N,

Ginsberg MH, et al. Robo4 stabilizes the vascular network by inhibiting pathologic

angiogenesis and endothelial hyperpermeability. Nature Medicine. 2008;14(4):448-

453.

135. Kawasaki T, Kitsukawa T, Bekku Y, Matsuda Y, Sanbo M, Yagi T, Fujisawa H. A

requirement for neuropilin-1 in embryonic vessel formation. Development.

1999;126(21):4895-4902.

136. Airas L, Niemela J, Salmi M, Puurunen T, Smith DJ, Jalkanen S. Differential

regulation and function of CD73, a glycosyl-phosphatidylinositol-linked 70-kD

adhesion molecule, on lymphocytes and endothelial cells. The Journal of Cell

Biology. 1997;136(2):421-431.

137. Sanchez M, Galy B, Muckenthaler MU, Hentze MW. Iron-regulatory proteins limit

hypoxia-inducible factor-2alpha expression in iron deficiency. Nature Structural &

Molecular Biology. 2007;14(5):420-426.

138. Urbich C, Dimmeler S. Endothelial progenitor cells: characterization and role in

vascular biology. Circulation Research. 2004;95(4):343-353.

139. Brantley-Sieders DM, Caughron J, Hicks D, Pozzi A, Ruiz JC, Chen J. EphA2

receptor tyrosine kinase regulates endothelial cell migration and vascular assembly

138

through phosphoinositide 3-kinase-mediated Rac1 GTPase activation. Journal of Cell

Science. 2004;117(Pt 10):2037-2049.

140. Adamo L, Naveiras O, Wenzel PL, McKinney-Freeman S, Mack PJ, Gracia-Sancho J,

Suchy-Dicey A, Yoshimoto M, Lensch MW, Yoder MC, Garcia-Cardena G, Daley

GQ. Biomechanical forces promote embryonic haematopoiesis. Nature.

2009;459(7250):1131-1135.

141. Hsiai TK, Wu JC. Hemodynamic forces regulate embryonic stem cell commitment to

vascular progenitors. Current Cardiology Reviews. 2008;4(4):269-274.

142. Wilson KD, Wu JC. The miR-145 pluripotency loop in human embryonic stem cells.

Cell Science. 2009;6(2).

143. Xu N, Papagiannakopoulos T, Pan G, Thomson J, Kosik K. MicroRNA-145 regulates

OCT4, SOX2, and KLF4 and represses pluripotency in human embryonic stem cells.

Cell. 2009;137(4):647-658.

144. Bartel DP. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell.

2004;116(2):281-297.

145. Carrington JC, Ambros V. Role of microRNAs in plant and animal development.

Science. 2003;301(5631):336-338.

146. Suh M-R, Lee Y, Kim JY, Kim S-K, Moon S-H, Lee JY, Cha K-Y, Chung HM, Yoon

HS, Moon SY, Kim VN, Kim K-S. Human embryonic stem cells express a unique set

of microRNAs. Developmental Biology. 2004;270(2):488-498.

147. Kanellopoulou C, Muljo SA, Kung AL, Ganesan S, Drapkin R, Jenuwein T,

Livingston DM, Rajewsky K. Dicer-deficient mouse embryonic stem cells are

defective in differentiation and centromeric silencing. Genes & development.

2005;19(4):489-501.

148. Murchison EP, Partridge JF, Tam OH, Cheloufi S, Hannon GJ. Characterization of

Dicer-deficient murine embryonic stem cells. Proceedings of the National Academy of

Sciences of the United States of America. 2005;102(34):12135-12140.

149. Wang Y, Medvid R, Melton C, Jaenisch R, Blelloch R. DGCR8 is essential for

microRNA biogenesis and silencing of embryonic stem cell self-renewal. Nature

Genetics. 2007;39(3):380-385.

150. Marson A, Levine SS, Cole MF, Frampton GM, Brambrink T, Johnstone S, Guenther

MG, Johnston WK, Wernig M, Newman J, Calabrese JM, Dennis LM, Volkert TL,

Gupta S, Love J, et al. Connecting microRNA genes to the core transcriptional

regulatory circuitry of embryonic stem cells. Cell. 2008;134(3):521-533.

151. Cordes KR, Sheehy NT, White MP, Berry EC, Morton SU, Muth AN, Lee TH, Miano

JM, Ivey KN, Srivastava D. miR-145 and miR-143 regulate smooth muscle cell fate

and plasticity. Nature. 2009;460(7256):705-710.

152. Gunaratne PH. Embryonic stem cell microRNAs: defining factors in induced

pluripotent (iPS) and cancer (CSC) stem cells? Current Stem Cell Research and

Therapy. 2009.

153. van Laake LW, Passier R, Doevendans PA, Mummery CL. Human embryonic stem

cell-derived cardiomyocytes and cardiac repair in rodents. Circulation Research.

2008;102(9):1008-1010.

154. Lakshmipathy U, Love B, Goff LA, Jornsten R, Graichen R, Hart RP, Chesnut JD.

MicroRNA expression pattern of undifferentiated and differentiated human embryonic

stem cells. Stem Cells and Development. 2007;16(6):1003-1016.

155. Catalucci D, Gallo P, Condorelli G. MicroRNAs in cardiovascular biology and heart

disease. Circulation: Cardiovascular Genetics. 2009;2:402-408.

156. Callis T, Pandya K, Seok H, Tang R-H, Tatsuguchi M, Huang Z-P, Chen J-F, Deng Z,

Gunn B, Shumate J, Willis M, Selzman C, Wang D-Z. MicroRNA-208a is a regulator

139

of cardiac hypertrophy and conduction in mice. Journal of Clinical Investigation.

2009;119(9):2772-2786.

157. van Rooij E, Sutherland LB, Liu N, Williams AH, McAnally J, Gerard RD,

Richardson JA, Olson EN. A signature pattern of stress-responsive microRNAs that

can evoke cardiac hypertrophy and heart failure. Proceedings of the National

Academy of Sciences of the United States of America. 2006;103(48):18255-18260.

158. van Rooij E, Sutherland LB, Qi X, Richardson JA, Hill J, Olson EN. Control of stress-

dependent cardiac growth and gene expression by a microRNA. Science.

2007;316(5824):575-579.

159. Care A, Catalucci D, Felicetti F, Bonci D, Addario A, Gallo P, Bang M-L, Segnalini

P, Gu Y, Dalton ND, Elia L, Latronico MVG, Hoydal M, Autore C, Russo MA, et al.

MicroRNA-133 controls cardiac hypertrophy. Nature Medicine. 2007;13(5):613-618.

160. Zhao Y, Ransom JF, Li A, Vedantham V, von Drehle M, Muth AN, Tsuchihashi T,

McManus MT, Schwartz RJ, Srivastava D. Dysregulation of cardiogenesis, cardiac

conduction, and cell cycle in mice lacking miRNA-1-2. Cell. 2007;129(2):303-317.

161. Chen J-F, Mandel EM, Thomson JM, Wu Q, Callis TE, Hammond SM, Conlon FL,

Wang D-Z. The role of microRNA-1 and microRNA-133 in skeletal muscle

proliferation and differentiation. Nature Genetics. 2006;38(2):228-233.

162. Ivey KN, Muth A, Arnold J, King FW, Yeh R-F, Fish JE, Hsiao EC, Schwartz RJ,

Conklin BR, Bernstein HS, Srivastava D. MicroRNA regulation of cell lineages in

mouse and human embryonic stem cells. Cell Stem Cell. 2008;2(3):219-229.

163. Gao X, Gulari E, Zhou X. In situ synthesis of oligonucleotide microarrays.

Biopolymers. 2004;73(5):579-596.

164. Bolstad BM, Irizarry RA, Astrand M, Speed TP. A comparison of normalization

methods for high density oligonucleotide array data based on variance and bias.

Bioinformatics. 2003;19(2):185-193.

165. Liang Y, Ridzon D, Wong L, Chen C. Characterization of microRNA expression

profiles in normal human tissues. BMC Genomics. 2007;8(1):166.

166. Baskerville S, Bartel D. Microarray profiling of microRNAs reveals frequent

coexpression with neighboring miRNAs and host genes. RNA. 2005;11(3):241-247.

167. Reiser PJ, Portman MA, Ning XH, Schomisch Moravec C. Human cardiac myosin

heavy chain isoforms in fetal and failing adult atria and ventricles. American Journal

of Physiology. Heart and Circulatory Physiology. 2001;280(4):H1814-H1820.

168. Sluijter JPG, van Mil A, van Vliet P, Metz CHG, Liu J, Doevendans PA, Goumans

M-J. MicroRNA-1 and -499 regulate differentiation and proliferation in human-

derived cardiomyocyte progenitor cells. Arteriosclerosis, Thrombosis, and Vascular

Biology. 2010;30:859-868.

169. van Rooij E, Quiat D, Johnson B, Sutherland L, Qi X, Richardson J, Kelm R, Olson E.

A family of microRNAs encoded by myosin genes governs myosin expression and

muscle performance. Developmental Cell. 2009;17(5):662-673.

170. Cohen E, Tian Y, Morrisey E. Wnt signaling: an essential regulator of cardiovascular

differentiation, morphogenesis and progenitor self-renewal. Development.

2008;135(5):789-798.

171. Yang B, Lin H, Xiao J, Lu Y, Luo X, Li B, Zhang Y, Xu C, Bai Y, Wang H, Chen G,

Wang Z. The muscle-specific microRNA miR-1 regulates cardiac arrhythmogenic

potential by targeting GJA1 and KCNJ2. Nature Medicine. 2007;13(4):486-491.

172. Luo X, Lin H, Pan Z, Xiao J, Zhang Y, Lu Y, Yang B, Wang Z. Down-regulation of

miR-1/miR-133 contributes to re-expression of pacemaker channel genes HCN2 and

HCN4 in hypertrophic heart. The Journal of Biological Chemistry.

2008;283(29):20045-20052.

140

173. Liu H, Kohane I, Kohane I. Tissue and process specific microRNA-mRNA co-

expression in mammalian development and malignancy. PloS One. 2009;4(5):e5436-

e5436.

174. Aoi T, Yae K, Nakagawa M, Ichisaka T, Okita K, Takahashi K, Chiba T, Yamanaka

S. Generation of pluripotent stem cells from adult mouse liver and stomach cells.

Science. 2008;321(5889):699-702.

175. Maherali N, Sridharan R, Xie W, Utikal J, Eminli S, Arnold K, Stadtfeld M,

Yachechko R, Tchieu J, Jaenisch R, Plath K, Hochedlinger K. Directly reprogrammed

fibroblasts show global epigenetic remodeling and widespread tissue contribution.

Cell Stem Cell. 2007;1(1):55-70.

176. Nakagawa M, Koyanagi M, Tanabe K, Takahashi K, Ichisaka T, Aoi T, Okita K,

Mochiduki Y, Takizawa N, Yamanaka S. Generation of induced pluripotent stem cells

without Myc from mouse and human fibroblasts. Nature Biotechnology.

2008;26(1):101-106.

177. Okita K, Ichisaka T, Yamanaka S. Generation of germline-competent induced

pluripotent stem cells. Nature. 2007;448(7151):313-317.

178. Park I-H, Zhao R, West JA, Yabuuchi A, Huo H, Ince TA, Lerou PH, Lensch MW,

Daley GQ. Reprogramming of human somatic cells to pluripotency with defined

factors. Nature. 2008;451(7175):141-146.

179. Wernig M, Meissner A, Foreman R, Brambrink T, Ku M, Hochedlinger K, Bernstein

BE, Jaenisch R. In vitro reprogramming of fibroblasts into a pluripotent ES-cell-like

state. Nature. 2007;448(7151):318-324.

180. Brambrink T, Foreman R, Welstead GG, Lengner CJ, Wernig M, Suh H, Jaenisch R.

Sequential expression of pluripotency markers during direct reprogramming of mouse

somatic cells. Cell Stem Cell. 2008;2(2):151-159.

181. Stadtfeld M, Maherali N, Breault DT, Hochedlinger K. Defining molecular

cornerstones during fibroblast to iPS cell reprogramming in mouse. Cell Stem Cell.

2008;2(3):230-240.

182. Jaenisch R, Young R. Stem cells, the molecular circuitry of pluripotency and nuclear

reprogramming. Cell. 2008;132(4):567-582.

183. Yamanaka S. Strategies and new developments in the generation of patient-specific

pluripotent stem cells. Cell Stem Cell. 2007;1(1):39-49.

184. Mikkelsen TS, Hanna J, Zhang X, Ku M, Wernig M, Schorderet P, Bernstein BE,

Jaenisch R, Lander ES, Meissner A. Dissecting direct reprogramming through

integrative genomic analysis. Nature. 2008;454(7200):49-55.

185. Laurent LC, Chen J, Ulitsky I, Mueller F-J, Lu C, Shamir R, Fan J-B, Loring JF.

Comprehensive microRNA profiling reveals a unique human embryonic stem cell

signature dominated by a single seed sequence. Stem Cells. 2008;26(6):1506-1516.

186. Chen C, Ridzon D, Lee C-T, Blake J, Sun Y, Strauss WM. Defining embryonic stem

cell identity using differentiation-related microRNAs and their potential targets.

Mammalian Genome. 2007;18(5):316-327.

187. Tang F, Kaneda M, O'Carroll Dn, Hajkova P, Barton SC, Sun YA, Lee C,

Tarakhovsky A, Lao K, Surani MA. Maternal microRNAs are essential for mouse

zygotic development. Genes & Development. 2007;21(6):644-648.

188. Sinkkonen L, Hugenschmidt T, Berninger P, Gaidatzis D, Mohn F, Artus-Revel CG,

Zavolan M, Svoboda P, Filipowicz W. MicroRNAs control de novo DNA methylation

through regulation of transcriptional repressors in mouse embryonic stem cells.

Nature Structural & Molecular Biology. 2008;15(3):259-267.

189. Zovoilis A, Nolte J, Drusenheimer N, Zechner U, Hada H, Guan K, Hasenfuss G,

Nayernia K, Engel W. Multipotent adult germline stem cells and embryonic stem cells

141

have similar microRNA profiles. Molecular Human Reproduction. 2008;14(9):521-

529.

190. Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis and display of

genome-wide expression patterns. Proceedings of the National Academy of Sciences

of the United States of America. 1998;95(25):14863-14868.

191. Landgraf P, Rusu M, Sheridan R, Sewer A, Iovino N, Aravin A, Pfeffer S, Rice A,

Kamphorst AO, Landthaler M, Lin C, Socci ND, Hermida L, Fulci V, Chiaretti S, et

al. A mammalian microRNA expression atlas based on small RNA library sequencing.

Cell. 2007;129(7):1401-1414.

192. Ge X, Wu Q, Wang SM. SAGE detects microRNA precursors. BMC Genomics.

2006;7:285-285.

193. Krueger BJ, Jeronimo Cl, Roy BB, Bouchard A, Barrandon C, Byers SA, Searcey CE,

Cooper JJ, Bensaude O, Cohen EA, Coulombe B, Price DH. LARP7 is a stable

component of the 7SK snRNP while P-TEFb, HEXIM1 and hnRNP A1 are reversibly

associated. Nucleic Acids Research. 2008;36(7):2219-2229.

194. Hayashita Y, Osada H, Tatematsu Y, Yamada H, Yanagisawa K, Tomida S, Yatabe

Y, Kawahara K, Sekido Y, Takahashi T. A polycistronic microRNA cluster, miR-17-

92, is overexpressed in human lung cancers and enhances cell proliferation. Cancer

Research. 2005;65(21):9628-9632.

195. Croce CM, Calin GA. miRNAs, cancer, and stem cell division. Cell. 2005;122(1):6-7.

196. Mendell JT. miRiad roles for the miR-17-92 cluster in development and disease. Cell.

2008;133(2):217-222.

197. Ventura A, Young AG, Winslow MM, Lintault L, Meissner A, Erkeland SJ, Newman

J, Bronson RT, Crowley D, Stone JR, Jaenisch R, Sharp PA, Jacks T. Targeted

deletion reveals essential and overlapping functions of the miR-17 through 92 family

of miRNA clusters. Cell. 2008;132(5):875-886.

198. Bentwich I, Avniel A, Karov Y, Aharonov R, Gilad S, Barad O, Barzilai A, Einat P,

Einav U, Meiri E, Sharon E, Spector Y, Bentwich Z. Identification of hundreds of

conserved and nonconserved human microRNAs. Nature Genetics. 2005;37(7):766-

770.

199. Reinhart BJ, Slack FJ, Basson M, Pasquinelli AE, Bettinger JC, Rougvie AE, Horvitz

HR, Ruvkun G. The 21-nucleotide let-7 RNA regulates developmental timing in

Caenorhabditis elegans. Nature. 2000;403(6772):901-906.

200. Viswanathan SR, Daley GQ, Gregory RI. Selective blockade of microRNA processing

by Lin28. Science. 2008;320(5872):97-100.

201. Fraser JK, Wulur I, Alfonso Z, Hedrick MH. Fat tissue: an underappreciated source of

stem cells for biotechnology. Trends in Biotechnology. 2006;24(4):150-154.

202. Judson R, Babiarz J, Venere M, Blelloch R. Embryonic stem cell-specific microRNAs

promote induced pluripotency. Nature Biotechnology. 2009;27(5):459-461.

203. Goff L, Davila J, Swerdel M, Moore J, Cohen R, Wu H, Sun Y, Hart R. Ago2

immunoprecipitation identifies predicted microRNAs in human embryonic stem cells

and neural precursors. PloS One. 2009;4(9):e7192-e7192.

204. Stadtfeld M, Nagaya M, Utikal J, Weir G, Hochedlinger K. Induced pluripotent stem

cells generated without viral integration. Science. 2008;322(5903):945-949.

205. Okita K, Nakagawa M, Hyenjong H, Ichisaka T, Yamanaka S. Generation of mouse

induced pluripotent stem cells without viral vectors. Science. 2008;322(5903):949-

953.

206. Kaji K, Norrby K, Paca A, Mileikovsky M, Mohseni P, Woltjen K. Virus-free

induction of pluripotency and subsequent excision of reprogramming factors. Nature.

2009;458(7239):771-775.

142

207. Soldner F, Hockemeyer D, Beard C, Gao Q, Bell G, Cook E, Hargus G, Blak A,

Cooper O, Mitalipova M, Isacson O, Jaenisch R. Parkinson's disease patient-derived

induced pluripotent stem cells free of viral reprogramming factors. Cell.

2009;136(5):964-977.

208. Woltjen K, Michael I, Mohseni P, Desai R, Mileikovsky M, Hmlinen R, Cowling R,

Wang W, Liu P, Gertsenstein M, Kaji K, Sung H-K, Nagy A. piggyBac transposition

reprograms fibroblasts to induced pluripotent stem cells. Nature. 2009;458(7239):766-

770.

209. Yu J, Hu K, Smuga-Otto K, Tian S, Stewart R, Slukvin II, Thomson JA. Human

induced pluripotent stem cells free of vector and transgene sequences. Science.

2009;324(5928):797-801.

210. Munz C, Bickham KL, Subklewe M, Tsang ML, Chahroudi A, Kurilla MG, Zhang D,

O'Donnell M, Steinman RM. Human CD4(+) T lymphocytes consistently respond to

the latent Epstein-Barr virus nuclear antigen EBNA1. The Journal of Experimental

Medicine. 2000;191(10):1649-1660.

211. Zhou H, Wu S, Joo J, Zhu S, Han D, Lin T, Trauger S, Bien G, Yao S, Zhu Y,

Siuzdak G, Schler H, Duan L, Ding S. Generation of induced pluripotent stem cells

using recombinant proteins. Cell Stem Cell. 2009;4(5):381-384.

212. Kim D, Kim C-H, Moon J-I, Chung Y-G, Chang M-Y, Han B-S, Ko S, Yang E, Cha

K, Lanza R, Kim K-S. Generation of human induced pluripotent stem cells by direct

delivery of reprogramming proteins. Cell Stem Cell. 2009;4(6):472-476.

213. Chen Z-Y, He C-Y, Ehrhardt A, Kay MA. Minicircle DNA vectors devoid of bacterial

DNA result in persistent and high-level transgene expression in vivo. Molecular

Therapy. 2003;8(3):495-500.

214. Chen Z-Y, He C-Y, Kay M. Improved production and purification of minicircle DNA

vector free of plasmid bacterial sequences and capable of persistent transgene

expression in vivo. Human Gene Therapy. 2005;16(1):126-131.

215. Chin M, Mason M, Xie W, Volinia S, Singer M, Peterson C, Ambartsumyan G,

Aimiuwu O, Richter L, Zhang J, Khvorostov I, Ott V, Grunstein M, Lavon N,

Benvenisty N, et al. Induced pluripotent stem cells and embryonic stem cells are

distinguished by gene expression signatures. Cell Stem Cell. 2009;5(1):111-123.

216. Aasen T, Raya A, Barrero M, Garreta E, Consiglio A, Gonzalez F, Vassena R, Bilic J,

Pekarik V, Tiscornia G, Edel M, Boue S, Izpisua Belmonte JC. Efficient and rapid

generation of induced pluripotent stem cells from human keratinocytes. Nature

Biotechnology. 2008;26(11):1276-1284.

217. Graf T, Stadtfeld M. Heterogeneity of embryonic and adult stem cells. Cell Stem Cell.

2008;3(5):480-483.

218. Huang S. Reprogramming cell fates: reconciling rarity with robustness. BioEssays.

2009;31(5):546-560.

219. Chang H, Oh P, Ingber D, Huang S. Multistable and multistep dynamics in neutrophil

differentiation. BMC Cell Biology. 2006;7:11-11.

220. Marchetto MCN, Yeo G, Kainohana O, Marsala M, Gage F, Muotri A.

Transcriptional signature and memory retention of human-induced pluripotent stem

cells. PloS One. 2009;4(9):e7076-e7076.

221. Miura K, Okada Y, Aoi T, Okada A, Takahashi K, Okita K, Nakagawa M, Koyanagi

M, Tanabe K, Ohnuki M, Ogawa D, Ikeda E, Okano H, Yamanaka S. Variation in the

safety of induced pluripotent stem cell lines. Nature Biotechnology. 2009;27(8):743-

745.

143

222. Bauters C, Lamblin N, Mc Fadden EP, Van Belle E, Millaire A, de Groote P.

Influence of diabetes mellitus on heart failure risk and outcome. Cardiovascular

Diabetology. 2003;2:1-1.

223. Olefsky JM. Treatment of insulin resistance with peroxisome proliferator-activated

receptor gamma agonists. The Journal of Clinical Investigation. 2000;106(4):467-472.

224. Vamecq J, Latruffe N. Medical significance of peroxisome proliferator-activated

receptors. The Lancet. 1999;354(9173):141-148.

225. Nissen SE, Wolski K. Effect of rosiglitazone on the risk of myocardial infarction and

death from cardiovascular causes. The New England Journal of Medicine.

2007;356(24):2457-2471.

226. Heishi M, Ichihara J, Teramoto R, Itakura Y, Hayashi K, Ishikawa H, Gomi H, Sakai

J, Kanaoka M, Taiji M, Kimura T. Global gene expression analysis in liver of obese

diabetic db/db mice treated with metformin. Diabetologia. 2006;49(7):1647-1655.

227. Keen HL, Ryan MJ, Beyer A, Mathur S, Scheetz TE, Gackle BD, Faraci FM,

Casavant TL, Sigmund CD. Gene expression profiling of potential PPARgamma

target genes in mouse aorta. Physiological Genomics. 2004;18(1):33-42.

228. Parton LE, Diraison F, Neill SE, Ghosh SK, Rubino MA, Bisi JE, Briscoe CP, Rutter

GA. Impact of PPARgamma overexpression and activation on pancreatic islet gene

expression profile analyzed with oligonucleotide microarrays. American Journal of

Physiology. Endocrinology and Metabolism. 2004;287(3):E390-E404.

229. Makino H, Miyamoto Y, Sawai K, Mori K, Mukoyama M, Nakao K, Yoshimasa Y,

Suga S-i. Altered gene expression related to glomerulogenesis and podocyte structure

in early diabetic nephropathy of db/db mice and its restoration by pioglitazone.

Diabetes. 2006;55(10):2747-2756.

230. Albrektsen T, Frederiksen KS, Holmes WE, Boel E, Taylor K, Fleckner J. Novel

genes regulated by the insulin sensitizer rosiglitazone during adipocyte differentiation.

Diabetes. 2002;51(4):1042-1051.

231. Rosen ED, Spiegelman BM. PPARgamma : a nuclear regulator of metabolism,

differentiation, and cell growth. The Journal of Biological Chemistry.

2001;276(41):37731-37734.

232. Marx N, Schonbeck U, Lazar MA, Libby P, Plutzky J. Peroxisome proliferator-

activated receptor gamma activators inhibit gene expression and migration in human

vascular smooth muscle cells. Circulation Research. 1998;83(11):1097-1103.

233. Kelly DP. PPARs of the heart: three is a crowd. Circulation Research.

2003;92(5):482-484.

234. Aasum E, Hafstad AD, Severson DL, Larsen TS. Age-dependent changes in

metabolism, contractile function, and ischemic sensitivity in hearts from db/db mice.

Diabetes. 2003;52(2):434-441.

235. Semeniuk LM, Kryski AJ, Severson DL. Echocardiographic assessment of cardiac

function in diabetic db/db and transgenic db/db-hGLUT4 mice. American Journal of

Physiology. Heart and Circulatory Physiology. 2002;283(3):H976-H982.

236. Belke DD, Larsen TS, Gibbs EM, Severson DL. Altered metabolism causes cardiac

dysfunction in perfused hearts from diabetic (db/db) mice. American Journal of

Physiology. Endocrinology and Metabolism. 2000;279(5):E1104-E1113.

237. Buchanan J, Mazumder PK, Hu P, Chakrabarti G, Roberts MW, Yun UJ, Cooksey

RC, Litwin SE, Abel ED. Reduced cardiac efficiency and altered substrate

metabolism precedes the onset of hyperglycemia and contractile dysfunction in two

mouse models of insulin resistance and obesity. Endocrinology. 2005;146(12):5341-

5349.

144

238. Yue P, Arai T, Terashima M, Sheikh AY, Cao F, Charo D, Hoyt G, Robbins RC,

Ashley EA, Wu J, Yang PC, Tsao PS. Magnetic resonance imaging of progressive

cardiomyopathic changes in the db/db mouse. American Journal of Physiology. Heart

and Circulatory Physiology. 2007;292(5):H2106-H2118.

239. Boden G. Role of fatty acids in the pathogenesis of insulin resistance and NIDDM.

Diabetes. 1997;46(1):3-10.

240. Sidell RJ, Cole MA, Draper NJ, Desrois M, Buckingham RE, Clarke K.

Thiazolidinedione treatment normalizes insulin resistance and ischemic injury in the

zucker Fatty rat heart. Diabetes. 2002;51(4):1110-1117.

241. Nesto RW, Bell D, Bonow RO, Fonseca V, Grundy SM, Horton ES, Le Winter M,

Porte D, Semenkovich CF, Smith S, Young LH, Kahn R. Thiazolidinedione use, fluid

retention, and congestive heart failure: a consensus statement from the American

Heart Association and American Diabetes Association. Diabetes Care.

2004;27(1):256-263.

242. Carley AN, Semeniuk LM, Shimoni Y, Aasum E, Larsen TS, Berger JP, Severson

DL. Treatment of type 2 diabetic db/db mice with a novel PPARgamma agonist

improves cardiac metabolism but not contractile function. American Journal of

Physiology. Endocrinology and Metabolism. 2004;286(3):E449-E455.

243. Dargie HJ, Hildebrandt PR, Riegger GnAJ, McMurray JJV, McMorn SO, Roberts JN,

Zambanini A, Wilding JPH. A randomized, placebo-controlled trial assessing the

effects of rosiglitazone on echocardiographic function and cardiac status in type 2

diabetic patients with New York Heart Association Functional Class I or II Heart

Failure. Journal of the American College of Cardiology. 2007;49(16):1696-1704.

244. Gilde AJ, van der Lee KAJM, Willemsen PHM, Chinetti G, van der Leij FR, van der

Vusse GJ, Staels B, van Bilsen M. Peroxisome proliferator-activated receptor (PPAR)

alpha and PPARbeta/delta, but not PPARgamma, modulate the expression of genes

involved in cardiac lipid metabolism. Circulation Research. 2003;92(5):518-524.

245. Bishop-Bailey D. Peroxisome proliferator-activated receptors in the cardiovascular

system. British Journal of Pharmacology. 2000;129(5):823-834.

246. Liu LS, Tanaka H, Ishii S, Eckel J. The new antidiabetic drug MCC-555 acutely

sensitizes insulin signaling in isolated cardiomyocytes. Endocrinology.

1998;139(11):4531-4539.

247. Duan SZ, Ivashchenko CY, Russell MW, Milstone DS, Mortensen RM.

Cardiomyocyte-specific knockout and agonist of peroxisome proliferator-activated

receptor-gamma both induce cardiac hypertrophy in mice. Circulation Research.

2005;97(4):372-379.

248. Rockman MV, Hahn MW, Soranzo N, Loisel DA, Goldstein DB, Wray GA. Positive

selection on MMP3 regulation has shaped heart disease risk. Current Biology.

2004;14(17):1531-1539.

249. Henney AM, Wakeley PR, Davies MJ, Foster K, Hembry R, Murphy G, Humphries S.

Localization of stromelysin gene expression in atherosclerotic plaques by in situ

hybridization. Proceedings of the National Academy of Sciences of the United States

of America. 1991;88(18):8154-8158.

250. Beyzade S, Zhang S, Wong Y-k, Day INM, Eriksson P, Ye S. Influences of matrix

metalloproteinase-3 gene variation on extent of coronary atherosclerosis and risk of

myocardial infarction. Journal of the American College of Cardiology.

2003;41(12):2130-2137.

251. Shimoni Y. Inhibition of the formation or action of angiotensin II reverses attenuated

K+ currents in type 1 and type 2 diabetes. The Journal of Physiology. 2001;537(Pt

1):83-92.

145

252. Dixon JE, Shi W, Wang HS, McDonald C, Yu H, Wymore RS, Cohen IS, McKinnon

D. Role of the Kv4.3 K+ channel in ventricular muscle. A molecular correlate for the

transient outward current. Circulation Research. 1996;79(4):659-668.

253. Lesage F, Lazdunski M. Molecular and functional properties of two-pore-domain

potassium channels. American Journal of Physiology. Renal, Fluid and Electrolyte

Physiology. 2000;279(5):F793-F801.

254. Chatterjee S, Kolmakova A, Miller M. The role of the phospholipid sphingomyelin in

heart disease. Current Opinion in Investigational Drugs. 2006;7(3):219-228.

255. Levade T, Augé N, Veldman RJ, Cuvillier O, Nègre-Salvayre A, Salvayre R.

Sphingolipid mediators in cardiovascular cell biology and pathology. Circulation

Research. 2001;89(11):957-968.

256. Unger RH, Orci L. Lipoapoptosis: its mechanism and its diseases. Biochimica et

Biophysica Acta. 2002;1585(2-3):202-212.

257. Baranowski M, Blachnio A, Zabielski P, Gorski J. Pioglitazone induces de novo

ceramide synthesis in the rat heart. Prostaglandins & Other Lipid Mediators.

2007;83(1-2):99-111.

258. Zhou YT, Grayburn P, Karim A, Shimabukuro M, Higa M, Baetens D, Orci L, Unger

RH. Lipotoxic heart disease in obese rats: implications for human obesity.

Proceedings of the National Academy of Sciences of the United States of America.

2000;97(4):1784-1789.

259. Hug C, Lodish HF. The role of the adipocyte hormone adiponectin in cardiovascular

disease. Current Opinion in Pharmacology. 2005;5(2):129-134.

260. Pineiro R, Iglesias M, Gallego R, Raghay K, Eiras S, Rubio J, Dieguez C, Gualillo O,

Gonzalez-Juanatey J, Lago F. Adiponectin is synthesized and secreted by human and

murine cardiomyocytes. FEBS Letters. 2005;579(23):5163-5169.

261. Natarajan R, Salloum FN, Fisher BJ, Kukreja RC, Fowler AA. Hypoxia inducible

factor-1 upregulates adiponectin in diabetic mouse hearts and attenuates post-ischemic

injury. Journal of Cardiovascular Pharmacology. 2008;51(2):178-187.

262. Chen MM, Ashley EA, Deng DXF, Tsalenko A, Deng A, Tabibiazar R, Ben-Dor A,

Fenster B, Yang E, King JY, Fowler M, Robbins R, Johnson FL, Bruhn L, McDonagh

T, et al. Novel role for the potent endogenous inotrope apelin in human cardiac

dysfunction. Circulation. 2003;108(12):1432-1439.

263. Iwanaga Y, Kihara Y, Takenaka H, Kita T. Down-regulation of cardiac apelin system

in hypertrophied and failing hearts: Possible role of angiotensin II-angiotensin type 1

receptor system. Journal of Molecular and Cellular Cardiology. 2006;41(5):798-806.

264. Lincoff AM, Wolski K, Nicholls SJ, Nissen SE. Pioglitazone and risk of

cardiovascular events in patients with type 2 diabetes mellitus: a meta-analysis of

randomized trials. Journal of the American Medical Association. 2007;298(10):1180-

1188.

265. Sears DD, Hsiao A, Ofrecio JM, Chapman J, He W, Olefsky JM. Selective

modulation of promoter recruitment and transcriptional activity of PPARgamma.

Biochemical and Biophysical Research Communications. 2007;364(3):515-521.

266. Guo L, Zhang L, Sun Y, Muskhelishvili L, Blann E, Dial S, Shi L, Schroth G, Dragan

YP. Differences in hepatotoxicity and gene expression profiles by anti-diabetic PPAR

gamma agonists on rat primary hepatocytes and human HepG2 cells. Molecular

Diversity. 2006;10(3):349-360.

146