Multimodal Biomarkers for Alzheimer’s disease
Candidates versus genomics and proteomics
Research collaborations relevant to biomarkers for AD:
• Proteome Sciences, Millipore Merck and GSK
• J&J and GE
• Precompetitive collaborative projects with multiple European Federation of
Pharmaceutical Industry Associations (EFPIA) partners
Other, non-biomarker, collaborations
• Astra Zeneca
• J&J
Consultancy / speaker fees
• Astra Zeneca
• Lundbeck
• Lilly
Conflict of interests
Karolinska
Johan Bengtsson
Tony Segerdahl
Christian Spenger
Eric Westman
King’s College
Simon Lovestone
Andy Simmons
Catherine Tunnard
University of Kuopio
Mervi Kononen
Hilkka Soininen
Ritva Vanninen
University of Perugia
Emanuela Costanzi
Patrizia Mecocci
Roberto Tarducci
Aristotle University
of Thessaloniki
Eleni Kantoglou
Penelope Mauredaki
Magda Tsolaki University of Lodz
Tadeusz Biegański
Iwona Kłoszewska
Radoslaw Magierski
University of Toulouse
Celine Caillaud
Pierre Payoux
Bruno Vellas
McGill University
Louis Collins
Alan Evans
Sebastian Muehlboeck
AddNeuroMed academic clinical and imaging team
NA-ADNI
J-ADNI
WW-ADNI
AIBL
C-ADNI AddNeuroMed
E-ADNI
Biomarkers for dementia – core pathology markers
Biomarkers for dementia – alternative approaches
-
Albumin
Pre-clinical studies Human studies
Biomarker discovery
Samples
Pre-clinical models
Clinical studies Validation
Samples
Biomarker discovery Validation
Human studies
Long term follow-on sample
Case control sample
Short term follow-on sample
Pro
teom
ics
Progression
markers
Diagnostic
markers
Surrogate markers
for use in trials
Bioinformatics
Intelligence
networks
Genom
ics
transcripto
mic
s
Candidate markers – hypothesis and informatics
Page 9
1.5 T sMRI and automated analysis
Regional cortical thickness – 34 areas
Regional cortical volume – 24 areas
36 cytokines measured by Luminex
commercial 30 plex and customised 6 plex
Analysis by Support Vector Machines to find the best
combination of predictors
Candidate markers – inflammatory proteins
Biomarkers of Mild Cognitive Impairment -
progression to dementia
Deborah Kronenberg, Simon Furney, Andreas Guentert, Andy Simmons
Total n= 48; 22 convertors Total n= 253 (99 convertors; 48 with imaging)
Intelligence networks to determine candidate
biomarkers
BioWisdom Sofia platform database access
Databases
Gene Ontology
NCBI Gene Expression Omnibus
Gensat Database
Diseases Database
HuGE Index – Tissue specific expression of mRNA database
KEGG – protein pathway information
Uni-Prot – database of proteins
On-line Mendelian Inheritance in Man – disease-associated genetic mutations
Biomarkers in AD knowledgebase (http://www.telemakus.net/AD/)
Textual Data
Medline
Text books (Adams and Victor’s Principles of Neurology)
Full text papers
Alzheimer Research Forum (http://www.alzforum.org/)
Essential Science Indicator’s (http://www.esi-topics.com/alzheimer/)
Workflow results
Proteins with a functional involvement
in defined pathological processes and
hallmarks of AD
26
Proteins/mRNAs expressed in AD
(defined brain regions)
5257
Proteins/mRNAs upregulated in AD
67
Candidates discovered through Intelligence Network
Candidates known to be involved in core pathology e.g:
• Apolipoprotein E
• BACE1
Candidates known to be involved in core pathology e.g:
• Clusterin
• Transthyretin
Candidates not previously thought of as markers e.g:
• Choline acetyltransferase
• Urokinase Plasminogen Activator Receptor (uPAR)
Validation process in uPAR and ChAT
Ines Greco, Julie Barnes, Niki Day
Discovery of markers based on disease hypothesis has proved successful
Discovery of markers beyond the core hypothesis has merit
Immune based markers show a clear but not consistent signal
Informatics based candidate discovery reveals novel markers
Summary – candidate markers
Candidates are limited by ‘what is known’ …….. genomics and proteomics are not
Blood based transcriptomic Markers
• Illumina Human HT-12 v3 Expression BeadChips (Illumina)
• N=332 (106 AD, 118 MCI [77 MCIStable, 41 MCIConverter] and 108 control
• Feature selection by random forests; t-tests with feature sets of 5, 10, 20, 50,
100 probes
• Random forests and SVM used as class prediction; combined with APOE and
MRI data and independently
Angela Hodges, Katie Lunnon, Richard Dobson
Blood based transcriptomic Markers – AD vs controls
Blood based transcriptomic Markers – AD vs controls
Blood based transcriptomic Markers
• Oxidative phosphorylation (OXPHOS) ; mitochondrial complexes I to V
• Transcription (GO:0006350; P = 9.1x10-147) / Translation (GO:0006412; P = 2.1x10-116)
• RNA Splicing (GO:0008380; P = 8.2 x10-102) / Processing (GO:0006396; P = 6.5 x10-98)
• ATP Binding (GO:0005524; P = 3.2 x10-94) / Protein Catabolism (GO:0044257 P = 1.7 x10-89)
• Ribosome Biogenesis (GO:0042254 P = 7.6 x10-37)
• Microtubule Cytoskeleton (GO:0015630; P = 1.8x10-33)
• Vesicle-mediated Transport (GO:0016192; P = 2.8 x10-28)
• Apoptosis (GO:0006915; P = 1.7 x10-7) and regulation (GO:0042981; P = 5.2 x10-13)
• Immune response and adhesion molecules
Gel based biomarker discovery in plasma
M.Wt
14 kDa
220 kDa
-
pH 3 Non-linear IPG pH 10
Albumin
IgG g chain
IgG l and k chain
Abdul Hye, Madhav Thambisetty, Latha Velayudhan
Limitations of gel based proteomics:
- Dynamic range is challenging
- To deplete or not to deplete ?
Case - control study
- two dimensional gel electrophoresis (2-DGE)
126.1 127.1 128.1 129.1 130.1
Study Reference Pool n=80
131.1
Plasma S1
Excise the complete
gel lane into ~ 10
individual eppendorfs
Reduce/Alkylate
+ trypsin
Discovery Proteomics Experiment using Isobaric Tandem
Mass Tags (TMT)
Quantitation
Plasma S2 Plasma S3 Plasma S4 Plasma S5
23
Proteomics Workflow Dec. 9th 2011
-
1.0
0.6
0.2
1.4
1.8
Rela
tive
op
tica
l d
en
sity
P<0.05
0.0 0.2 0.4 0.6 0.8 1.0
1 - Specificity
0.0
0.2
0.4
0.6
0.8
1.0
Sen
sit
ivit
y
ROC Curve
Complement Factor H (CFH)
n>500
Validation of CFH as a marker
CFH biomarker replication
Cutler et al (2008) Proteomics Clin Appl
Gel based separation, depletion, MS identification
47 x 47 (cases x controls)
CFH identified as predominant marker on 2DGE
Akuffo et al (2008) Biomarkers 13 618-636
Rosiglitazone AD trial
Gel based proteomics at baseline and 24 weeks of therapy
CFH shows correlation with response to rosiglitazone
Mueller et al. (2010) J Alzheimers Dis ePub
CFH identified as altered in AD using proteomics
Correlation with cortical atrophy
Correlation with cognition (MMSE)
Correlation with speed of decline
Using endophenotypes to search for biomarkers
Protein ID - O
1Complement C3
2g -Fibrinogen
3Serum albumin
4Complement Factor-I
Clusterin
Clusterin
7Serum amyloid P-
component
8α1-microglobulin
Protein ID - O
1Complement C4-A
2γ -Fibrinogen
3Complement
component C8 g
chain
Clusterin
5Complement C4-A
6Complement C4-A
7Apolipoprotein A-I
8Apolipoprotein A-I
9Transthyretin
[s1]
Hippocampal atrophy
n=44 Progression
n=51
Madhav Thambisetty, Latha Velayudhan
Validation – correlation with imaging, cognition and progression
AD only : n=113,R= -0.31 and p=0.001
AD and MCI: n=220, R=-0.14 and p=0.04
Imaging: entorhinal atrophy Cogntion : MMSE
AD only : n=576, r=-0.22; p<0.001
Andreas Guentert, Abdul Hye,
Anna Kinsey
Non-rapid
decline
Non-rapid
decline
Rapid
decline
Rapid
decline
Retrospective decline
n=344
Prospective decline
n=237
[Clu
ste
rin
] pla
sm
a
Progression: before and after sample point
Plasma Clusterin association with brain amyloid
In man….
Susan Resnick, Madhav Thambisetty and the
BLSA study team
W/T APP/PS1
In brain….
In mouse….
Plasma clusterin in life correlates with brain clusterin
in superior temporal gyrus
R=0.47 ; p = 0.027 ; N=22
David Howlett, Paul Francis, Andreas
Guentert
Muzamil Saleem, Andreas Guentert
September 6th 2009
N=14,000
N=16,000
Next steps – assay design for qualification
Intellectual property on ~30-protein panel protected by
KCL/Proteome sciences. Licensed for research use to Millipore
Joint development with Proteome Sciences and Millipore
Funding from MRC
Luminex xMAP panel
MS panel using Reaction Monitoring and isobaric tags
Validation / qualification study design
Validation / qualification workflow
AddNeuroMed
DCR
genADA
TOTAL
595 126 275 996
Normal 213 68 114 395
MCI 172 - - 172
AD 210 58 161 429
Normal MCI AD Sig.
n=395 n=172 n=429
Age, yrs 76.22
(±6.0, 54-93)
76.34
(±5.8, 65-90)
77.21
(±6.3, 58-96)
Non-sig
Sex (% female) 51.6% 49.4% 55.5% Non-sig
APOE genotype
(% e4)
27.1% 35.0% 59.9% P<0.001*
MMSE
28.92
(±1.26, 22-30)
26.91
(±2.9, 0-30)
21.15
(±5.2, 0-30)
P<0.001*
N=1000; validation study
Marker correlation with ADAS-cog
CFH CRP
Classifier scheme
AD vs. Control models
Train Test
Used variables Sens.
[%]
Spec.
[%]
Acc.
[%]
Sens.
[%]
Spec.
[%]
Acc.
[%]
1. Proteins 79.2 66.7 73.8 75.7 71.6 73.7
2. APOE, age,
gender
67.5 66.7 67.5 69.9 56.8 63.6
MCI vs. Control models
Train Test
Used variables MCI [%] CTL [%] Acc.
[%]
MCI [%] CTL [%] Acc.
[%]
1. Proteins 74.2 67.4 68.0 85.7 61.1 68.6
2. APOE, Age,
gender
59.7 61.8 60.2 19.1 72.6 56.2
AD vs. MCI models
Train Test
Used variables AD [%] MCI [%] Acc.
[%]
AD [%] MCI [%] Acc.
[%]
1. Proteins 75.3 72.6 75.9 71.9 66.7 70.4
2. APOE, age,
gender
71.4 54.8 65.7 92.2 14.3 69.7
Both candidate and data-driven approaches confirm a signature of potential
markers in blood
Candidate studies repeatedly find an inflammatory signal but the exact nature of
the signal differs between studies
Transcript based data-driven studies find a signature but further replication and
refinement is needed
Proteomics yields a consistent signature from discovery by endophenotypes,
through replication, biological confirmation and qualification studies
Summary
BRC Bioinformatics team
Martina Sattlecker
Anbarasu Lourdasamy
Simon Furney
Gerome Breen
Richard Dobson
KHP Biomarkers team
Abdul Hye
Joanna Riddoch-Contreras
Rufina Leung
Andreas Guentert
Mohammed Aiyaz
Madhav Thambisetty (NIA)
Katie Lunnon
Latha Velayudhan
Angela Hodges
Simon Lovestone
AddNeuroMed Imaging team
Eric Westman
Sebastian Muehlboeck
Sergi Costafreda
Lars-Olof Wahlund
Christian Spenger
Alan Evans
Andrew Simmons
Collaborators in Proteome Sciences, Millipore, GSK, GenADA, UCLA
AddNeuroMed collaborators
Supported by:
FP6 EU funding to AddNeuroMed
NIHR Biomedical Research Centre for Mental Health
at the South London and Maudsley NHS Foundation
Trust and King’s College London
MRC Centre for Neurodegeneration research
Additional funding from Alzheimer’s Research Trust