Professor Michael Levin's presentation at Meningitis Research Foundation's 2013 conference Meningitis & Septicaemia in Children & Adults www.meningitis.org/conference2013

1. The emerging picture of host
genetic control of meningococcal
disease: Insights from Genome
wide association studies
(GWAS)
Michael Levin
Imperial college London
MRF Conference November 2013
© Imperial College London

2. WHY Do some children
develop Meningococcal
Infection or other forms of
meningitis and not others ?
Is it an accident or in the
genes?

3. Epidemiology and Infection June 2003
Number of affected case/sibling
Sibling familial risk ratio
pairs
Duration between case/sibling MD onsets
35
30
25
20
12
10
11
AIM
8
Quantify the host genetic component of
6
6 meningococcal disease susceptibility
15 4
10 2
1
4
0
Method
1
7
1
5 0
All data< 1 > 1 w eek 1 > 1 month > 3 >3 mths 6 months > 9 months > 12 months
>
>1 mth- months >
6
>
> 12
Calculate the co-efficient of>familial 9
week week-<
<3
<
clustering: 6 R mths < mths < mths
Duration between MD onsets in case/affected sibling pairs
1 mth
mths
mths 9 mths 12 mths
© Imperial College London

4. Why undertake genetic studies of
Infectious diseases?
• Identify fundamental pathways causing
susceptibility
• Identify those susceptible
• Identify key components of
pathophysiology
• Explain individual differences in clinical
presentation or outcome
• Identify targets for therapy
© Imperial College London

5. © Imperial College London

6. Sites for genetic influence
Nasopharyngeal colonisation
Invasion of bloodstream
Survival in blood
Asymptomatic
Bacteraemia
Meningitis
Septic Shock
Purpura
Fulminans
© Imperial College London

7. Susceptibility versus severity genes
Susceptibility
Severity
Control colonisation invasion
Survival in blood
Determine inflammatory
response
Extent of coagulopathy
Organ failure
Susceptibility genes
present with greater
frequency in cases vs
controls
Severity genes may have
same frequency in cases
and controls preferentially
in severe/fatal cases
© Imperial College London

8. How to identify these genes?
• Candidate gene approach
– Biology-based, mice & men
– Limited to what you know
• Genome-wide association
– Can detect ‘novel’ genes
– Use SNP arrays
• Linkage analysis
– Small family studies
– Not population based
• Next-gen sequencing
– Whole exome
– Whole genome?

9. The St Mary's / Imperial College approach 1995+
Detailed clinical phenotype
Immunology, biochemistry, microbiology, haematology
Candidate genes based on pathophysiology
Case control association study
Confirm in second
cohort
Family study
Other
cohorts
Protein expression biological studies
Identify gene defects
© Imperial College London

10. © Imperial College London
Lots of Candidate
gene Studies

11. Problems with candidate gene genetic
association studies
•Innumerable candidates?
•Only tests what is known
•Many published studies methodologically
flawed
• Few replication studies
•Inadequate correction for multiple hypothesis
testing, genetic admixture
© Imperial College London

13. The feasibility of Genome-Wide Association
Studies
• Now possible to type
thousands to millions of
SNPS across entire genome
in each individual
• Cost of genotyping
decreasing
Requires Large patient and control cohorts
High throughput technology and
Bioinformatics expertise
© Imperial College London

14. HapMap
A program to chart genetic variation
within the human genome
• Single neucleotide
polymorphisms (SNPs)scattered
across genome
10 million sequence
variants in each individual

17. Exponential increase in Validated GWAS findings

18. Genome wide study of
meningococcal disease
• An ESPID collaboration
• MRF support for establishment
of St Mary’s/ UK Cohort 19952009 and on-going analysis

19. Meningococcal GWAS
Sonia Davila
Martin Hibberd
Chui Chin Lim
DNA QC:
Chang Hua
Wong
Dennis Tan
Jie Wen Tay
Computing:
Jieming
Chen
Mike Levin
Victoria Wright
Jan Hazelzet
David Inwald
Taco Kuijpers
Marieke Emonts Simon Nadel
Willemijn Breunis
Helen Betts
Lachlan Coin
Enitan Carrol
Ronald de Groot
Peter Hermans
Werner Zenz
Alexander Binder
University of Santiago de
Compostela (Spain)
Federico Martinon-Torres
Antonio Salas
© Imperial College London

20. Study design of the MRF/ ESPID GWAS
UK Caucasian MD cases
UK Caucasian controls
Genome-Wide Association Study using
Illumina 610 Hap-quad chip
Discovery
study
Bioinformatic Analysis
Top ‘hits’ & selection of
candidate genes
Validation in European cohorts (Austria, Holland, Spain ) using Illumina SQNM
Replication
study
Bioinformatic Analysis
Fine-mapping using Illumina ISelect
Functional Studies
Gene Expression/Proteomics
Identification of Key Pathways
© Imperial College London

21. 2010 Sep;42(9):772-6.

22. UK Meningococcal GWAS
-log (observed P-values)
How do we detect significance ?
-log (expected P-values)

23. Manhattan plot of significant SNPs in
meningococcal disease patients
Martinón-Torres, ESPID 2012

24. Factor H & Factor H related protein region on Chr.1

25. Replication:
Austrian/Dutch & Spanish cohorts
Davila Nature genetics 2010

26. Factor H/ FH related protein region
• Multiple SNPs spanning FH- FHR protein
region
• Highly significant in all 3 cohorts
• Definitive identification of FH and FHR as
influencing susceptibility

27. Complement Factor H binds to Meningococcal
protein (FhbP)
Fh on endothelial
cells bind
meningococci
Inhibits C3b
Reduces killing

28. What is the role of FH related
proteins?
• Sequence homology to FH
• Individual variation in ratio of FH/FHR
proteins 1-4 ?
• Competition between FH and FHR for binding
to Meningococcal FHBP ?
• Further work to define mechanisms at a
functional and protein level

29. Further exploration of meningococcal genentics
EU Childhood Life-threatening Infectious Disease Study

31. Staged programme to identify genetic basis of Meningococcal disease
Vaccine GWAS
UK GWAS
Spanish GWAS
Central Europe GWAS
Prospective recruitment
of MD & other
bacterial infection
Identify top Hits
Replicate in 2nd
vaccine cohort
Cross-validation between European cohorts
Meta-analysis of combined UK, Spanish, CE European GWAS
Fine mapping &
sequencing of candidate
regions
Genotype/RNA/Protein
studies [Functional studies]
Evaluation of variants in
animal models
Validation in prospective MD & other bacterial infection cohorts [EU & Africa]
Genomic analysis of Extreme Phenotype cohorts
Predictive biomarkers of susceptibility & severity & clinical translation
Severity
analysis /
Pathway
analysis

32. Meningococcal disease and age-related macular
degeneration share genetic susceptibility loci
ESIGEM network and EUCLIDS consortium
ABCA4 top hit
after CFH region
• ABCA4 encodes a
protein expressed in
retinal
photoreceptors.
• Mutations in ABCA4
are associated with
degenerative
macular diseases.
Martinón-Torres, ESPID 2012

33. AMD
MD
Leading cause of blindness
in western societies
Leading bacterial threat
In children
ABCA4
CFH
Complement
mediated
pathogenesis
????
Martinón-Torres, ESPID 2012

34. EUCLIDS
• Large sample size enables meta analysis of
the European cohort to identify genes
controlling severity and outcome
• Pathway based analysis

35. Multiple sites for genetic regulation of
Inflammatory response to a pathogen
© Imperial College London

36. Pathway Based analysis of Genome
wide studies
• Can the combined effect of mutiple variants
in a biological pathway be used for analysis
of GWAS and gene expression data ?

37. Test for pathway association
1. Select pathway
•
•
Create a gene list
Map all SNPs to genes within 10K
2. Max single-SNP trend test statistic of the 4 genetic models
(for every SNP in the pathway)
Contigency table of case-control genotype SNP data
aa
aA
AA
Total
Cases
r0
r1
r2
R
Controls
s0
s1
s2
S
Total
n0
n1
n2
N
H 0 : pi
aa
aA
AA
Additive
0
1
2
Dominant
0
1
0
0
1
Heterozygote
0
1
0
0,1, 2
*
T
U
1
U
N
1/ 2
[varH 0 (U )]

var H 0 (U )
N
2
RS
0
3
2
x i ( Sri
2
2
2
i
[
R si )
i 0
x ni
2
(
xi ni ) ]
1
Recessive
qi , i
Cochran-Armitage test statistic2
Z
4 genetic models1
1.
2.
Null hypothesis
Sasieni, Biometrics, 1997
Freidlin et al, Hum Hered, 2002
N
i 0
i 0
Max statistic
C A M A X = m ax (C A | x=(0,1,2) , C A | x=(0,1,1) , C A | x=(0,0,1) , C A | x=(0,1,0) )

38. Variable selection and cross validation identifies
Genes with consistent association

39. Visualization of genomic risk
Individual genomic fingerprint Type 1
diabetes
1
0.5
0
Predicted probability of disease
SNPs in T1D logistic model of disease
Real disease status
a
b
c
Individuals

40. SEVERITY AND PATHWAY
ANALYSIS OF EUCLIDS
MENINGOCOCCAL COHORTS
• UK GWAS 475 cases/ 4703 controls
• SPANISH GWAS 417 cases/882 controls
• AUSTRIAN/DUTCH GWAS
344 cases/2557 controls
End Points : Death; amputations :skin
grafting; mechanical ventilation;severity
scores

41. Intermediate phenotypes

44. Death associated with second messenger signaling

45. Second messenger
signaling

46. Death and severity controlled by
second messenger signalling
The second messenger signalling systems are used by all cells to
switch on and off multiple processes
Environment
Processes controlled
by SMS
Genetic variation in the second messanger on/off switch
may control intensity of inflammation in MD

47. Amputations and skin loss

50. Severity and pathway analysis of
EUCLIDS GWAS is revealing complex
control of phenotypes by genes within
each individual
• Further validation of initial findings
planned in new EUCLIDS cohort
• New Pathways as therapeutic targets

51. Acknowledgements
MD patients and families
Control families
Sonia Davila
Martin Hibberd
Chui Chin Lim
DNA QC:
Chang Hua Wong
Dennis Tan
Jie Wen Tay
Taco Kuijpers
Willemijn Breunis
Enitan Carrol
Genotyping:
Wee Yang Meah
Khai Koon Heng
Sigeeta
Ronald de Groot
Peter Hermans
Rajaram
Computing:
Kar Seng Sim
Jieming Chen
Jan Hazelzet
Marieke Emonts
Victoria Wright
David Inwald
Simon Nadel
Helen Betts
Lachlan Coin
Harieta Eleftherohorinou
University of Santiago de
Compostela (Spain)
Werner Zenz
Alexander Binder
Federico Martinon-Torres
Antonio Salas
© Imperial College London

52. Thank You!