Transcript Life or Cell Death: Deciphering c-Myc Regulated Gene Networks In

Life or Cell Death:
Deciphering c-Myc Regulated Gene
Networks In Two Distinct Tissues
Sam Robson
MOAC DTC, Coventry House, University of Warwick,
Gibbet Hill Road, Coventry, CV4 7AL
Mco
A
Outline
1. Introduction to c-Myc
2. Transgenic in vivo models –
skin versus pancreas
3. Methods
4. Results
5. Generalised linear models
Project Aims
•
Using two distinct switchable in vivo cMyc models, we aim to:
– Analyse differences in gene-expression
– Identify c-Myc regulated genes in cell
replication and cell death
– Improve understanding of complex c-Myc
activity in diseases such as cancer
– To understand how and why c-Myc can
regulate vastly different paradoxical
phenotypes in vivo
1: Introduction to c-Myc
– Transcription factor involved in wide range
of cellular functions – “Dual function”
– May regulate up to 15% of all genes
– Deregulated in majority of human cancers
– Therapeutic target?
– Exact mechanisms not well understood –
we know WHAT c-Myc does, but we want
to know WHY it does it
– In vitro studies miss complex interactions
of surrounding environment on cell fate
c-Myc Regulated Processes
Growth
External
Signals
c-Myc
Proliferation
(eg. mitogens,
survival factors)
Apoptosis
Loss of Differentiation
Cell-Cycle Progression
Gene Activation
MYC MAX
Ub
CCND2
CDK4
Cyclin D2 CDK4
CACGTG
CUL1
CKS
p27KIP1
Proteosome
p27KIP1
E-Box sequence in
promoter sequence
of target gene
p27KIP1
Cyclin E
CDK2
CAK
Inactive
Cyclin E
CDK2
Active
P
MIZ-1
MYC MAX
p15Ink4b (CDKN2B)
p27 (not known if Miz-1 is required)
Sp1/Sp3
MYC
p15Ink4b (CDKN2B)
p21Waf1 (CDKN1A)
Apoptosis – Cell Death
FAS Ligand
FAS “Death
Receptor”
Death Induced Signalling Complex (DISC)
FADD
BID
BCL-2
Apoptosome
Procaspase 8
Caspase
Cascade
Effector
ARF
caspases
FLIP
BAX/
BAK
tBID
APAF-1
MOMP
c-Myc
Procaspase 9
Cytochrome c
Smac
DIABLO
Mitochondrion
ATP
IAPs
BIM
IAPs
PUMA
Apoptosis
p53
AIF
NOXA
Endo G
Cellular
targets
Omi/
Htra2
Effector caspases
2: Transgenic in vivo models
– Controlled activation of c-Myc functions in
target cells
– Can analyse immediate effects of c-Myc
activation
– Targetted to pancreatic islet β-cells (insulin
promoter) and skin supra-basal keratinocytes
(involucrin promoter)
– Activation of c-Myc can lead to drastically
different phenotypes – Replication in skin,
apoptosis in pancreas
Inactive
Active
MycERTAM
HAT TRRAP
Transgenic Model – c-MycERTAM
RNA
Polymerase
Legend
Myc Box I
Myc Box II
Basic
Helix-Loop-Helix
Leucine Zipper
Estrogen Receptor
Myc-Max
complex
Transformationbinds
E-box
TRRAP
recruits a histone
Transcription
domain
4-Hydroxytamoxifen
Myc
sequence
ofProtein
acetyltransferase
(HAT). This
Associated
Max Max
binds
Myc
target
gene
acetylates
(TRRAP)
binds to nucleosomal
atMBII
leucine
histones
resulting in
with
help from
helix-loop-helix
chromatin remodelling,
MBI
4-OHT
binds
zipper region
allowing access by RNA
Bound
Heat
estrogen
receptor
HSP90
TAM
Polymerase
for gene
ER
ShockupProtein
CACGTG opening
bHLHz90
transcription
domain.
TAM
c-MycER
Activation
Inactive
Skin
Active
Suprabasal
layer
Pancreas
Pelengaris et al. (1999), Molecular Cell, Vol. 3(5), 565-577
Pelengaris et al. (2002), Cell, Vol. 109(3), 321-334
Suprabasal
layer
TAM
c-MycER
Activation
• Skin
Unchecked proliferation, no
apoptosis - Replication
• Pancreas
Synchronous cell cycle entry
and apoptosis – Death
• Myc activation regulates
two opposing phenotypes
Pelengaris et al. (1999), Molecular Cell, Vol. 3(5), 565-577
Pelengaris et al. (2002), Cell, Vol. 109(3), 321-334
3: Methods
– Microarrays – High throughput technique
– “Transcriptomics” – Analysis at mRNA
level
– LCM to ensure RNA homogeneity
– mRNA very delicate! Degradation by
RNAses
– Huge amount of work to develop robust
protocol for extraction of RNA of suitable
quality and yield from LCM
– Many technical problems to overcome
Workflow
1: Treatment of
Transgenics
2: Extraction of
Tissue
Controlled activation of
c-Myc in two diverse
tissues
6: Microarray
Hybridisation
Hybridise cRNA to
microarrays
3: Laser
Capture
Microdissection
Excision of target tissue
QC
5: 2-Cycle IVT
Preparation of cRNA for
microarray hybridisation
Isolation of homogenous
tissue
QC
4: mRNA
Extraction
Isolate mRNA from target
cells
QC
7: Microarray
Data Analysis
8: Validation
Studies
9: Functional
Validation
Analysis of microarray
data
Validation studies to
confirm results
Linking results to the
biology of the system
Experimental Setup
Untreated with 4-OHT
x3
4
8
16
32
4
x3
8
16
32
x3
Time course
Gene
Expression
Gene
Expression
Time course
Pancreas
Tissue
x3
Time course
Gene
Expression
Gene
Expression
Time course
Skin
Tissue
Treated with 4-OHT
4
8
16
32
4
8
16
32
Laser Capture Microdissection
• Heterogeneity of tissue may cause
problem with in vivo studies
• β-cells make up only ~2% of
pancreas
• LCM allows isolation of
homogenous cell populations
• Optimisation of protocol for LCM of
islets – No other protocols
available
• LCM of skin not possible – too
tough
Laser Capture Microdissection
1: Find Islet
2: Cut Islet
3: Lift Islet
4: Extracted Islet
Laser Capture Microdissection
1: Find Islet
2: Cut Islet
3: Lift Islet
4: Extracted Islet
Laser Capture Microdissection
1: Find Islet
2: Cut Islet
3: Lift Islet
4: Extracted Islet
Technical problems
• mRNA very unstable – Great care taken to
prevent degradation
• Pancreas is notorious for being full of RNAses!
• Standard LCM protocols very long –
Optimisation of suitable protocol for islets
• Small mRNA yield from LCM
• Logistics of 84 samples – Lots of preparation!
• Batching of samples – Randomisation to prevent
systematic errors and batching effects
• ~1 year for LCM optimisation
~9 months from tissue to microarray results!
RNA Integrity
Poor quality:
Okay quality:
Majority of peaks
at lower levels
18S and 28S peaks
more prominent, but
many peaks at
lower levels
Good quality:
Excellent quality:
Fewer peaks at
lower levels
18S and 28S peaks
clear with almost no
peaks at lower levels
Effect of RNA Quality on Yield
RNA Yield vs Quality
100
90
Starting cRNA
ug
80
70
60
50
40
30
20
10
0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
RIN
• General trend between RNA quality (RIN) and yield (Starting
cRNA)
• Only 1 low starting cRNA samples below RIN=5 cutoff
• Implies RIN may not be a great estimator of overall RNA
yield
Effect of RNA Quality on Yield
RNA Yield vs Quality
Skin
100
90
Starting cRNA
ug
80
Pancreas
70
60
50
40
30
20
10
0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
RIN
• In general, skin samples have higher RNA quality and yield
than pancreas samples
• Many differences between skin and pancreas
– Greater number of ribonucleases in pancreas
– Homeostasis maintained in skin
– More intense processing for pancreas tissue RNA compared to skin
Microarray Analysis
• Each feature measures one 25mer nucleotide sequence.
• 25-mer sequence specifically binds
biotin labelled cRNA.
• Hundreds of identical 25mers per
feature.
• Fluorescence readings give relative
mRNA concentration - gene expression
• 11-20 features per gene.
• Very, very expensive!
Courtesy of Affymetrix - www.affymetrix.com
4: Results
– Quality control of microarray data – Several
outliers but generally good quality data
– Outliers increase variance – Remove for
differential analysis
– Outliers spread nicely amongst conditions –
importance of randomisation!
– Analysis of early time points – Direct c-Myc
targets
Skin vs Pancreas
Skin
Pancreas
• Clustering – Group
similar samples
together
• Branching tree like
structure – samples
on the same branch
most similar
• Data cluster nicely on
tissue (some outliers)
• Given the protocol,
the data looks great!
Gene Expression Analysis
Pancreas
Skin
• Differential Expression
Look for genes with changing
expression across conditions
• Statistics
Compare distributions
between conditions to look
for significant changes
• Error
Biological error, technical
error, random error
• Functional Analysis
Similar expression profile
implies related biological
mechanisms
Tissue-Specific Differentiation Markers
Insulin
~4-fold down
in pancreas
Involucrin
~2-fold down
in skin
Cell-Cycle Progression
Cyclin D2
~2-fold up E
Cyclin
in skin
~4-fold up
in pancreas
• Ccnd2 and CDK4 upregulated in skin –
Indicates G1/S cell cycle progression
• No change in pancreas – Odd
• CDK inhibitor p27 downregulated in both
• Cyclin E upregulated in pancreas and
not skin – Again, very odd
CDK4
~4-fold up
in skin
p27KIP1
~2-fold down in pancreas
~4-fold down in skin
Apoptosis
p19ARF
~2-fold up
in pancreas
•
Increase in p19 – Oncogenic stress (p53
dependent pathway)
•
No change in p53 at transcriptional level –
Changes may occur at protein level
•
Massive increase in Fas receptor
expression – Extrinsic pathway
•
Myc seems to drive apotosis through
extrinsic and intrinsic pathways
Fas Receptor
~6-fold up
in pancreas
p53
No change
5: Generalised Linear Models
– Most microarray studies focus on one
or two main parameters
– Multi-factorial approach poses
problems with significance analysis
– Use of generalised linear models
– Widely applicable particularly for
clinical studies
– Collaboration with Agilent –
Implementation in Genespring GX
Generalised Linear Models
• Unsupervised linear regressive technique.
• Model gene expression data as a linear
combination of parameter variables:
y  b1 x1  b2 x2  ...  b p x p  
y = (y1,…,yn)T is the response variable (gene expression) for each sample
xi = (x1,…,xn)T are the explanatory variables (1 ≤ i ≤ p) for each sample
bi is the model coefficient for explanatory variable xi
n is the number of samples, p is the number of parameters
ε is some error term
Generalised Linear Models
• Can be used in the following ways:
1. To check how much of an effect other
parameters have on gene expression (eg
batching effects)
2. To find genes that change based on particular
parameters while taking other parameters and
interactions into account (eg clinical data)
• Makes fewer assumptions of data distribution
• Works with unbalanced experiment designs –
useful for clinical data.
Generalised Linear Models
• Program written in statistical programming
language R
• Written as part of the Bioconductor project
• Implemented in GeneSpring GX (Agilent) – Aim to
translate into JAVA for complete integration
• Close collaboration with Agilent
• Currently testing the program on a number of
diverse data sets
• MOAC (Shameless plug) – First crop of interdisciplinary scientists almost ready
Further Work
• Analysis of microarray data – Cluster analysis,
differential analysis, network analysis, etc.
• Use of GLM algorithm and comparison of results
with standard methods (ANOVA)
• Validation of results – Immunohistology,
quantitative real time PCR, etc.
• Functional validation – siRNA, ChIP-on-chip, etc.
• Translation of GLM program to JAVA for
implementation in GeneSpring GX version 8
Conclusion
• c-Myc regulates replication and cell death
• Web of pathways to decipher – Tissue
context in vivo
• Seems to initiate apoptosis through
combination of extrinsic and intrinsic
pathways
• Want to find the ‘suicide note’ for the
pancreas – why choose death?
Acknowledgements
Project Supervisors:
Michael Khan
David Epstein
Stella Pelengaris
Advisory Committee:
Robert Old
Manu Vatish
James Lynn
Special thanks:
Helen Bird
Lesley Ward
Sue Davis
Heather Turner
Ewan Hunter
Sponsors:
EPSRC, BBSRC, AICR, Eli Lilly and Amylin Pharmaceuticals Inc.
Mco
A
Acknowledgements
Luxian
David
Stella
Mike
Vicky
Sevi
Sylvie