This document summarizes a study that analyzed gene expression and regulation in adipose tissue from obese and non-obese individuals. MicroRNA expression was found to be different between the two groups, with many miRNAs downregulated in obesity. One miRNA in particular, miR-193b, was shown to regulate secretion of the inflammatory factor CCL2. Motif activity response analysis identified transcription factors with significantly different activity between obese and non-obese individuals. Together, the results provide new insights into the perturbed transcriptional regulation of adipogenesis and inflammation in human obesity.

1. Sequencing the transcriptome
reveals complex levels of regulation
Carsten O. Daub
Copenhagenomics 2012-06-15

2. RIKEN – Omics Science Center (OSC)

3. Karolinska Institute
Science for Life Laboratory – SciLifeLab Department of Biosciences and Nutrition
Technology Hospital

4. Outline
• Characterize cells on the molecular level
• Finding regulators of gene expression
– We observe gene expression
– What are the regulatory elements causing the
observed changes in gene expression?
• Example: Inflammation of adipocytes in
obesity

5. Cap Analysis of Gene Expression method
RNA extraction
RNA
5’
Gene
CAGE library preparation
3’
1. CAP trapper 遺伝子
Gene Gene
Genome
2. Trehalose extension method
3. CAGE library RNA
RNA
CAP
Sequence
Tag sequencing with the next-generation sequencer
1. Genome Sequencer FLX (Roche/454)
2. Genome Analyzer System (illumina/ Solexa)
3. SOLiD (Applied Biosystems)
Data processing
Quality control
Statistical variation of the obtained sequence
Extraction of tag sequences
Clustering
Mapping
Statistical variation of the mapping result
Visualization with “genome browser”
Statistical analysis T. Shiraki et al, PNAS, 100, 15776-15781 (2003)

6. Traditional CAGE method
RNA extraction
CAGE library preparation CGCATGGTCGATAGACTTG
1. CAP trapper
2. Trehalose extension method
3. CAGE library GTGCGCGTCGAATATCGAT
Sequence CGAATATCGATAGACTTG
Tag sequencing with the next generation sequencer
1. Genome Sequencer FLX (Roche/454)
2. Genome Analyzer System (illumina/ Solexa)
3. SOLiD (Applied Biosystems)
Data processing
Quality control
Statistical variation of the obtained sequence
Extraction of tag sequences
Clustering
Mapping
Statistical variation of the mapping result
Visualization with “genome browser”
Statistical analysis
T. Shiraki et al, PNAS, 100, 15776-15781 (2003)

7. Traditional CAGE method
RNA extraction
Genome
CAGE library preparation
1. CAP trapper
2. Trehalose extension method
3. CAGE library
Sequence
Tag sequencing with the next generation sequencer
1. Genome Sequencer FLX (Roche/454)
2. Genome Analyzer System (illumina/ Solexa)
3. SOLiD (Applied Biosystems)
Data processing GTGCGCGTCGAATATCGAT
Quality control
Statistical variation of the obtained sequence
Extraction of tag sequences
Clustering
Mapping
Statistical variation of the mapping result
Visualization with “genome browser”
Statistical analysis

8. CAGE – Promoter Types
Nature reviews Genetics, 8(6), 424–436. doi:10.1038/nrg2026

9. FANTOM4 – A Systems Approach
Monoblast-like THP-1 cells were stimulated by PMA to differentiate them into monocyte-like cells.
10 time point samples were collected during differentiation.
Monoblast-like Monocyte-like
0 1 2 4 6 12 24 48 72 96 hour
PMA
Replicates Microarray check
Deep CAGE
RIKEN1
RIKEN3 TF qRT-PCR
RIKEN5 Not good
RIKEN6
Illumina (47K probes) miRNA microarray
10 time points

10. Motif Activity Response Analysis (MARA)
Promoter Analysis & Motif Activity
29,857 promoters were identified.
Out of these promoters 23,403 were associated with 9026 genes.
Genome eps
CAGE tag
Promoter1 m1 m1 m1 m2 m3
PMA Promoter2 m1 m4
・・・・
0h 1h 2h 4h 6h 12h 48h 72h 96h
Promoter
29,857 m1 m5
Number of CAGE tags that
mapped on the same site
e ps m
R pm Ams
Reaction efficiency
• Number of possible binding sites Effective concentration
•
THP-1 cells are a Degree of conservation cell line which upon PMA treatment can differentiate into an
monoblastic leukemia of the motif
•
adherent monocyte like cell (CD14+, CSF1R+)
Chromatin status
Suzuki, Forrest, van Nimwegen et al. Nature Genetics 2009, 41:5

11. Transcription regulation network consisting of 30 core motifs (F4)
55 out of 86 edges were supported by experiments/in the literature. (Novel prediction works well!!)
Enriched GO: from cell growth related to cell function related Motif activity
Immune Inflammatory
response Up
Monocyte response
Cell adhesion Down
Transient
Size of nodes：
Significance of motifs
Edge support
Green: siRNA
Monoblast Microtubele Red: literature
Mitosis Cell cycle cytoskele Blue: ChIP
：enriched GO
for regulated
H. Suzuki et al. Nature Genetics, 41:5, 553-562 (2009) genes

12. FANTOM5
Static networks Time-courses
iPS
Stem ES
State Transition
basins
Endo Meso Ecto
Differentiated
states FANTOM4
FANTOM5

13. Defining a cell (F5)
•Surface markers
•Morphology (shape, volume, polarity)
•Single or multinucleated, enucleated
•Ploidy
•Motility (adherent, resident,
migratory)
•Differentiation potential
•Self renewal potential
•Developmental/lineage history
•Tissue of origin
•Developmental age (doublings?)
•Doubling time
Defined outputs (eg growth factors)
Response to inputs
Self reinforcing stable internal network

14. Defining a cell (F5)
Chr1 Chr2 Chr22 ChrX ChrY
•Surface markers
Human 1bp •Morphology (shape, volume, polarity)
3x109bp
genome •Single or multinucleated, enucleated
•Ploidy
All promoters on human genome will be revealed.
•Motility (adherent, resident,
migratory)
•Differentiation potential
The •Self renewal potential
most objective definition of the cell !!
•Developmental/lineage history
•Tissue of origin
•Developmental age (doublings?)
•Doubling time
Defined outputs (eg growth factors)
Response to inputs
Self reinforcing stable internal network Transcriptional regulatory NW

15. High reproducibility of Helicos-CAGE (FANTOM5)
Heliscope doesn't use PCR, then avoids amplification bias.
r=0.99 r=0.98 r=0.89 r=0.69
HeliScope CAGE

16. Promoters at unexpected area(1)
PUM2
In adipocyte, this is the only one promoter in this gene
NCOA1
In adipocyte, this is the only one promoter in this gene
The 2nd example has good EST evidence
Both are annotated in Gencode
Some orphans are NOT orphans

17. Expression profile (F5)
NKX2-5
This profile is useful
for biomarker development
Tissue-specific
Heart-fetal

18. Selfish DNA
• Transposable elements (TEs)
– the DNA sequence spreads by forming additional
copies of itself within the genome
– makes no specific contribution to the reproductive
success of its host organism.
– transposition can be "copy and paste" or "cut and
paste"
• Nobel prize in 1983 to Barbara McClintock

19. Expression of Repeat Elements
Mouse Human

20. Nat Genet. 2009 May;41(5):563-71.
Enrichment of TE near TSS
mouse human mouse human
Mapping bias Over-expression
Unrelated to expression when close to genes

21. FANTOM Collaborators, Thanks!
Australia Western Australian Institute for Medical Research
Peter KLINKEN, Louise WINTERINGHAM
Canada McGill University
Hisashi MIURA, Josee DOSTIE
The University of British Columbia, Center for Molecular Medecine and Therapeutics
Thomas Jonghyun HA
Denmark University of Copenhagen, Department of Biology
Robin ANDERSSON, Albin SANDELIN, Eivind VALEN
Finland University of Helsinki, Department of Medical Genetics
Alessandro BONETTI
France University Pierre & Marie Curie, Laboratoire Microorganisms Genomics
Hugues RICHARD
Germany Charité - Universitätsmedizin Berlin, Allergy Center
Magda BABINA
University Hospital Regensburg
Christian SCHMIDL, Michael REHLI
Italy Dulbecco Telethon Institute
Valerio ORLANDO, Beatrice BODEGA
Fondazione Bruno Kessler (FBK)
Marco CHIERICI, Cesare FURLANELLO, Marco RONCADOR
International School for Advanced Studies (SISSA)
Stefano GUSTINICH, Silvia ZUCCHELLI
National Lab of Italian Consortium for Biotechnology (L.N.C.I.B.)
Silvano PIAZZA, Claudio SCHNEIDER, Roberto VERARDO
Thank you.
Japan Database Center for Life Science
Hidemasa BONO
Keio University, School of Medicine
Shigeo KOYASU, Kazuyo MORO, Jun-ichi FURUSAWA General Organizer
Kyushu University
Daisuke SUGIYAMA Yoshihide HAYASHIZAKI
National Institute of Advanced Industrial Science and Technology (AIST), Computational Biology Research Center
(CBRC)

23. Transcriptional regulation of
adipogenesis is perturbed in obesity

24. Obesity – Molecular basis
• Medical condition with excess of body fat
• Increased fat mass is associated with a low-grade
inflammatory state in white adipose tissue
• Adipocytes secrete key inflammatory proteins
– TNFa, IL6, CCL2 (MCP1)
– Markedly increased expression in obesity
• CCL2 attracts macrophages from the blood
stream
• Mechanism of CCL2 expression control poorly
understood

25. Social aspects
• Increases the likelihood of various diseases
– heart disease, type 2 diabetes, obstructive sleep
apnea, certain types of cancer, and osteoarthritis
• The medical care costs of obesity in the United
States in 2008 totaled about US$147 billion
(Health Affairs 2009; 28(5): w822-w831)

26. Goals
• Find regulators of key obesity genes in man
– Substantial differences in adipose tissue gene
expression between obese / non-obese (JCEM, 2005)
– Includes many genes involved in the functional
regulation of adipose tissue
• Adipogenesis, metabolism, adipokine production,
inflammation
• The regulatory elements that control obesity key
genes are so far poorly characterized
• TFs and upstream modulators (epigenetics, miRNAs)

27. Study design
Cohort 1
Cohort 2
Obese (n=30) and
Obese and non-obese
non-obese (n=26)
women (n=12)
women
1. Hypothesis generation 3. Validation & mechanistical studies
In vitro
Intact adipose Isolated, mature
2. Relevance differentiated
tissue fat cells
adipocytes
4. Physiological role
Gene expression Gene expression
miRNA expression Knockdowns
qRT-PCR of Over-expressions
TFs & miRNAs 3’ UTR assay

28. Cohort 1

29. miRNA expression
• 20 differentially expressed candidates identified
in WAT (from cohort 1)
• Some of them previously observed in obese mice
– Most miRNAs down-regulated in obese samples
• Fits well with the observation that 90% of differentially
regulated genes are up-regulated
– Differential expression validated with qPCR in WAT
• 14 miRNAs validated in tissue and purified cells
• In mature differentiated fat cells (from cohort 2)
– All 20 expressed
– 11 miRNAs diff. expressed and overlapping with WAT

30. miRNA – functional study
CCL2 secretion
• CCL2 (MCP1) secretion
• Over-expressing 12 miRNAs in pre-adipocyte
differentiation
CCL2
IL6
3.0 *** TNF
**
2.5
2.0
** **
(microRNA/control)
***
Relative secretion
1.5
1.0
*** * * **** ** ** ** ***** *
*** *** ***
*** *** *** *** ** ***
0.5 *** *** ***
*** ***
0
Control
Let-7d
Let-7a
miR-16
miR-126
miR-145
miR-143
miR-652
miR-193b
miR-92a
miR-26a
miR-486-5p
miR-342-3p

31. Gene expression
• Substantial expression differences between obese
and non-obese human individuals in WAT (cohort
1)
– Many genes differentially expressed
– Around 100 transcription factors
– 90% of genes up-regulated
– Confirms previous findings
• Known up/down-regulated genes
• GO terms
– Up: inflammation, extracellular matrix
– Down: metabolism, insulin signaling

32. Obesity regulation
• Motif Activity Response Analysis (MARA)
a) Transcription factor binding site (TFBS) motifs
with significant difference in activity between
obese and non-obese, intersected with
b) targets with significant difference in expression
according to SAM (5% FDR)
→ nodes in candidate “core” network

33. Principal Component Analysis
on Motif Activities
Obese 55
0.15
Non-obese
Second principal component
Body mass index (kg/m2)
0.10 45
0.05
35
0
-0.05 25
-0.10
15
-0.15
-0.15 -0.10 -0.05 0 0.05 0.10 0.15
-0.25 -0.15 -0.05 0.05 0.15
Second principal component
First principal component
1400
160 80
Fat cell volume (pL)
1050 135 65
Body fat (%)
Waist (cm)
700 110 50
350 85 35
0 60 20
-0.15 -0.10 -0.05 0 0.05 0.10 0.15 -0.15-0.10-0.05 0 0.05 0.10 0.15 -0.15-0.10-0.05 0 0.05 0.10 0.15
Second principal component Second principal component
Second principal component

34. Regulatory network
• TF regulation from MARA
– Overlap obesity network and adipogenesis network
• miRNAs into network
– Based on target predictions (Miranda, TargetScan)
• Final candidate network
– Developed during adipocyte differentiation
– Perturbed in obesity
– Modulated by miRNAs
• Sub-network: inflammation
– Obesity is characterized by chronic, low-grade inflammation in
adipose tissue
• Coupled to obesity related conditions, i.e. insulin sensitivity
• TNF, IL6, CCL2

35. Sub-network validation
Knock/over-expression
→ mRNA, protein
Luciferace assay

36. Physiological evaluation
1.4 1400
0.9 1050
Fat cell volume (pL)
0.4 700
Log10 HOMAIR
-0.1 350
-0.6 0
-0.2 0.2 0.6 1.0 1.4 -0.2 0.2 0.6 1.0 1.4
Log10 miR-193b levels (A.U.) Log10 miR-193b levels (A.U.)
2.0 160
Waist circumference (cm)
1.4 135
Log10 insulin-stimulated
adipocyte lipogenesis
0.8 110
0.2 85
-0.4 60
-0.2 0.2 0.6 1.0 1.4 -0.2 0.2 0.6 1.0 1.4
Log10 miR-193b levels (A.U.) Log10 miR-193b levels (A.U.)

37. Conclusions
• Regulation of adipose tissue inflammation is
multilayered
– Mediated by miRNAs directly and indirectly through TFs
– Developed during pre-adipocyte differentiation
– Perturbed in obesity
• Approach is viable for exposing unknown, potentially
clinically relevant regulation
– Hypothesis free initial step enables study of other
processes such as adipogenesis, metabolism and adipokine
production
• Published in Diabetes on June 11, 2012

38. Acknowledgments
• Lipid Lab @ Karolinska Institutet
– Niklas Mejhert
– Peter Arner*
– Ingrid Dahlman
– Agné Kulyté
– Mikael Rydén
• Biozentrum Basel
– Erik van Nimwegen*
– Mikhail Pachkov
– Piotr Balwierz
• OSC RIKEN
– Erik Arner
– Yoshihide Hayashizaki
Grant from MEXT to YH.

39. Solving challenges in data analysis –
Conclusion
• Bring together technology and clinical research
– Understand experimental system and data in detail
– Understand the results of the computational analysis
– Jointly & iteratively
– Soft skills very helpful
• Aim for concrete findings that can be functionally
validated