the power of gene expression profiling to unravel behaviour cathy fernandes, jose paya-cano, frans...
DESCRIPTION
the interaction of multiple genes and their products a snapshot of the simultaneous gene expression across thousands of genes Microarrays, Mice & Behavioural Genetics mice are excellent models genetic overlap with humans differences in behaviour and gene expression genomic information access to fresh brain tissue nominate new candidate genes for behaviourTRANSCRIPT
The power of gene expression profiling to unravel behaviour
Cathy Fernandes, Jose Paya-Cano, Frans Sluyter, Ursula D'Souza, Robert Plomin,
Leonard C Schalkwyk
Social, Genetic and Developmental Psychiatry Centre
Institute of Psychiatry
King’s College London
• Hippocampal gene expression and cognitive ability
• Background
• Gene expression using the Affymetrix GeneChip system
• Hippocampal gene expression profiles across eight different inbred mouse strains
Outline
• the interaction of multiple genes and their products
• a snapshot of the simultaneous gene expression across thousands of genes
Microarrays, Mice & Behavioural Genetics
• mice are excellent models • genetic overlap with humans• differences in behaviour and gene expression • genomic information• access to fresh brain tissue
nominate new candidate genes for behaviour
Gene expression studies using microarrays
• Sandberg et al (2000) • six brain regions in 129SvEv and C57BL/6
• 24 genes strain-specific expression across all brain regions (240 genes regional gene expression differences)• re-analysis by Pavlidis and Noble (2001) identified many more genes with strain-specific (63 genes) and/or region-specific (600 genes) expression
• Gene expression profiles • during development (Mody et al, 2001)
• resulting from ageing (Jiang et al 2001, Lee et al, 2000)• behavioural manipulations (Leil et al, 2002)
• environmental manipulations (Rampon et al, 2000)
Hippocampal gene expression profiling across eight inbred mouse strains
AIMS
• determine how much gene expression is due to genetic variation
• to expand on the currently available gene expression data by increasing the number of mouse strains studied
• to find biologically relevant strain differences in gene expression, filtering out random individual differences
• to produce tightly controlled, replicated data • reliable pattern of gene expression • maximise detection of relatively small differences in expression
Selection of inbred strainsSelected from Group A of the Mouse Phenome Database
• commonly used strains with available genetic and phenotypic information• progenitors in transgenesis and mutagenesis studies
• progenitors of recombinant inbred, consomic and congenic strains A/J* BALB/cByJ C3H/HeJ DBA/2J*129S1/SvImJ*
C57BL/6J #
FVB/NJ SJL/J
• differ in activity, exploration, anxiety, learning, aggression
* Celera Mouse Genome Sequencing Projects# Public Mouse Genome Sequencing Projects
- key area of the brain involved in behaviours such as learning/memory and anxiety
- discrete area and is of a sufficient size in the mouse to allow a precise and highly reproducible dissection
- yield sufficient quantities of mRNA for microarray work
- strain-specific gene expression (Sandberg et al, 2000)
The role of the hippocampus
Procedure
- male mice (6 per strain, 48 mice in total) from Jackson Laboratories (USA) aged 5-6 weeks
- acclimatised in our barrier facility for 8 weeks (singly housed)
- killed by cervical dislocation, in a randomised order, aged 13-14 weeks (over 3 days to minimise any effect of time of day)
- hippocampus was immediately dissected out, snap frozen on dry ice and stored at –80 0C
- dissections done by the same operator and completed within 1 minute for each mouse
Procedure(contd)
The following procedures were carried out to minimise stress to the mouse prior to killing:
• minimal handling of mice
• transported to the procedure room in their home cage and killed within 3 minutes of transport
• method of kill
• killed by the same operator
Analysis
1. The data was analysed in parallel using Affymetrix MAS5 and Li and Wong PM-only model (dChip v1.2, Li and Wong 2001a)
• differ in methods used to summarise the probesets and for normalisation of the arrays
2. Signal values analysed in R (http://www.r-project.org/, Ihaka & Gentleman 1996, Neuwirth & Baier 2001)
- one-way ANOVA (results were filtered using a p value cut-off of 4 x 10-6 (p≤ 0.05 following Bonferroni correction for 12,488 probesets)
3. Hierarchical clustering (Eisen, 1998) was carried on the ANOVA filtered (p < 4 x 10-6) gene expression levels
- strain means for the probesets fit a normal distribution
- 252 (MAS5) and 200 (dChip) probesets with p values for difference of < 4x 10-6
- 100 probesets were identified in both analysis programs
- discrepant probes most commonly are those of low signal - many of the strain differences due to up or down-regulation, rather than presence or absence, of the transcript
- the bulk of the probesets expression profiles are very similar (pairwise correlations between chips MAS5: 0.894 - 0.997, dChip: 0.901 - 0.997)
Results
Clustering
- numerous and clear strain differences in gene expression
Clustering(contd)
(in 10 different random permutation runs of the strain factor to assess the false positive rate, only two p-values < 4 x 10-6 were found (i.e. 1000 fold fewer than with the real factor)
- strains cluster together
Clustering(contd)
- among clusters of probesets, several reunite multiple probesets representing the same transcript
- for example, four caspase 9 probesets cluster together (more highly expressed in BALB/cByJ and C3H/HeJ)
-many can be identified which are biologically plausible
- for example, one striking cluster includes 5 loci from the H2 region of chromosome 17: H2-d (3 probesets), H2-k, and Qa, which are expressed above the mean in FVB/NJ and DBA/2J (BUT does not correlate with the H2 haplotypes of the strains)
Effect of gene mutation on expression
- increased expression of Alad in DBA/2J compared to C57BL/6J strain (gene is present in two copies in DBA and one in C57BL/6J), Claudio et al (1997)
Alad (aminolevulinate, delta-, dehydratase)
0
50
100
150
200
250
300
129 A BALB C3H C57BL DBA FVB SJL
Strains
MAS
5 ex
pres
sion
leve
l
Effect of gene mutation on expression
- Gas5 gene is known to harbour mutations that affect the stability of its mRNA transcript in the 129 substrains (Muller et al 1998)
Gas5(growth arrest specific 5)
0
200
400
600
800
1000
129 A BALB C3H C57 DBA FVB SJL
Strains
Mas
5 ex
pres
sion
leve
l
Some potential candidates
- microtubule-associated protein tau has key structural functions and is essential to beta-amyloid-induced neurotoxicity- preliminary data on protein levels (Western blots) support the expression RNA data (D'Alcontres and Hanger, Neuroscience, IoP)
Mapt(microtubule-associated protein tau)
0
200
400
600
800
1000
1200
129 A BALB C3H C57 DBA FVB SJL
Strains
MAS
5 ex
pres
sion
leve
l
Pam (peptidylglycine alpha-amidating monooxygenase)
0.0
50.0
100.0
150.0
200.0
129 A BALB C3H C57BL DBA FVB SJL
Strains
MAS
5 ex
pres
sion
leve
l
- a key bifunctional enzyme in the activation of neuropeptides - gene maps to chromosome 1 at 57.5 cM (an ethanol-induced loss of righting reflex locus at chr 1, 43 and 59 cM)
Some potential candidates
Camk2a( calcium/calmodulin-dependent protein kinase II alpha)
0100200300400500600700
129 A BALB C3H C57 DBA FVB SJL
Strains
MAS
5 ex
pres
sion
leve
l
Some potential candidates
- Camk2a is implicated in the establishment of long-term potentiation (Bejar et al 2002) and spatial learning (Silva et al 1992, Giese et al 1998)
- BUT does not correlate with learning in these strains ?
Correlation• more and more phenotype data for inbred strains is
available
• it may be possible to find meaningful correlations with expression data (WebQTL)
• similar to Grupe et al in silico genetic mapping
• shortcomings also resemble Grupe
Aggression• Consensus aggression ranking (intermale offensive aggression), Sluyter:
FVB/NJ> SJL/J> BALB/cByJ> C3H/HeJ> DBA/2J> C57BL/6J> 129S1/SvImJ> A/J
• Spearman correlation with our chip data:
name rho pval probesetcatechol-O-methyltransferase 0.90 0.0020 98535_atfibroblast growth factor 1 0.93 0.0009 100494_at
COMT expression correlation
1 2 3 4 5 6 7 8
350
400
450
500
550
Strain ranks
CO
MT
- link between low COMT activity and increased aggression in mice and humans (Gogos et al, 1998; Lachman et al, 1998; Jones et al, 2001)
- biased towards detection of abundantly expressed, well- characterised genes
- rare transcripts, short half-life, alternative splicing
BUT low-abundance mRNAs or those expressed only at very specific times in development and/or processes may be key to determining the behavioural phenotype
Limitations
- cellular heterogeneity
- polymorphisms may obscure differences or create spurious ones
Results (contd)
• one third of the highly significant probesets have one or more additional probesets representing the same transcript
• compare or combine multiple probesetsMicrotubule-associated protein tau
02004006008001000120014001600
129 A BALB C3H C57 DBA FVB SJL
Strains
DCHI
P ex
pres
sion
leve
l
Caspase 9
020406080100120140160
129 A BALB C3H C57 DBA FVB SJL
Strains
MAS
5 ex
pres
sion
leve
l
Carbonic anhydrase 14
0
100
200
300
400
500
600
700
129 A BALB C3H C57 DBA FVB SJL
Strains
DCHI
P ex
pres
sion
leve
l
Multiple probesets
Schalkwyk et al 1999
History of inbred strains - analysis of CIDR data (http://www.cidr.jhmi.edu/) by Schalkwyk et al (1999)
SPRET/Ei
CAST/Ei
SKIVE/Ei
MOLF/Ei
MOLG/Dn100
9855
PERC/Ei
PERA/Rk100
C58/J
C57BR/cdJ
C57L/J100
C57BL/10J
RF/J
C57BL/6J100
100
80
100
BTBR+Ttf/t
LP/J
129T2/SvEm
129X1/SvJ51
129P3/J
129S2/SvPa
129S6/SvEv98
94
100
100
100
KK/HlJ
RIIIS/J30
NZW/LacJ
NZB/BlNJ99
MRL/MpJ
AKR/J74
BUB/BnJ
NOD/LtJ78
14
ST/bJ
NON/LtJ
SJL/J
SWR/J
FVB/NJ85
7558
34
9
A/J
BALB/cJ96
SM/J
P/J
BDP/J100
69
I/LnJ
DBA/2J
DBA/1J100
34
23
CE/J
CBA/J
CBA/CaJ88
SF/CamEi
C3H/HeJ
C3HeB/FeJ100
57
84
17
9
22
21
11
13
26
46
100
100
HS progenitor strains
Wagner parsimony analysis using MIX (Felsenstein 1988b) of microsatellite data (298 loci from all 19 autosomes and X) on 48 strains, transformed into binary characters according to Schalkwyk et al 1999, and using SPRET as outgroup. Internal figures are the number of bootstrap replicates out of 100 supporting each group. The overall topology agrees with Schalkwyk (1999) except that the C57 and 129 groups are reversed.
Cheverud’s take
Witmer et al 2003
New microsatellites
10
SM/J
NOD/LtJ74 ISS/IbgC57BL/6ByJ
C57BL/6J
100
35
ILS/Ibg
P/J
I/LnJ
78
27
DBA/2J
CBA/JC3H/HeJ
94
96
4513
NZB/BINJ
129X1/SvJBTBR
367 5
PL/J
AKR/J
RF/J
99
77
FVB/NJBUB/BnJ
81
23
RIIIS/JKK/HIJ
2341648
BALB/cByJ
BALB/cJ
100LP/J
A/J100
97
MI6i/Pomp
NON/LtJ
45
LG/J
NZW/Wehi
NZO/Wehi
94
30
SJL/JSWR/J
100
28
WSB/Ei
PERA/EiPERC/Ei
ZALENDE/Ei
TIRANO/Ei
100
CZECHII/Ei
PWK/Ph SKIVE/Ei
76
73
MOLF/Ei
JF/1
MSM
100
99
99
CASA/Ei
CAST/Ei
99
77
SPRET/Ei
PANCEVO/Ei
100
10045
3537
100
1884
PANCEVO/EiSPRET/EiCZECHII/EiPWK/PhSKIVE/Ei
7673
MOLF/EiJF/1MSM
10099
99
CASA/EiCAST/Ei
99
77
PERC/EiPERA/EiWSB/EiSM/JNOD/LtJ
74
ISS/IbgC57BL/6ByJC57BL/6J
10035
ILS/IbgP/JI/LnJ
7827
DBA/2JCBA/JC3H/HeJ
9496
45
13
NZB/BINJ129X1/SvJBTBR
367
5
PL/JAKR/JRF/J
9977
FVB/NJBUB/BnJ
81
23
RIIIS/JKK/HIJ
23
4
16
48
BALB/cByJBALB/cJ
100
LP/JA/J
10097
84
MI6i/PompNON/LtJ
45
18
LG/JNZW/WehiNZO/Wehi
9430
SJL/JSWR/J
100
28
100
37
35
45
ZALENDE/EiTIRANO/Ei
100
100
100
100
100
PANCEVO/EiSPRET/Ei
CZECHII/EiPWK/PhSKIVE/Ei
7673
MOLF/EiJF/1MSM
10099
99
CASA/EiCAST/Ei
99
77
PERC/EiPERA/Ei
WSB/EiSM/JNOD/LtJ
74
ISS/IbgC57BL/6ByJC57BL/6J
10035
ILS/IbgP/JI/LnJ
7827
DBA/2JCBA/JC3H/HeJ
9496
45
13
NZB/BINJ129X1/SvJBTBR
367
5
PL/JAKR/JRF/J
9977
FVB/NJBUB/BnJ
81
23
RIIIS/JKK/HIJ
23
4
16
48
BALB/cByJBALB/cJ
100
LP/JA/J
10097
84
MI6i/PompNON/LtJ
45
18
LG/JNZW/WehiNZO/Wehi
9430
SJL/JSWR/J
100
28
100
37
35
45
ZALENDE/EiTIRANO/Ei
100
100
100
100