/

University of Southern California Dissertations and Theses

/

Neuroimaging in complex polygenic disorders

(USC Thesis Other)

Neuroimaging in complex polygenic disorders

PDF

Transcript (if available)

Content

Neuroimaging in complex polygenic disorders

by

Daniel Allen Rinker

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(NEUROSCIENCE)

May 2018

Copyright 2018 Daniel Allen Rinker

i

Dedication

In loving memory of my sister, Coco.

ii
Acknowledgements
This project was funded by NIH grant U54 EB020403, additionally by National Multiple
Sclerosis Society grant RG4680A1/1, US National Institute of Child Health and Human
Development (RO1HD050735) and National Health and Medical Research Council
(Project Grants No. 496682 and 1009064).
I would like thank study participants of the Queensland Twin Imaging Study and the
Alzheimer’s Disease Neuroimaging Initiative.
I would like to express deep gratitude to my mentor, Paul M. Thompson, for all of the
support, advice and endless scientific opportunity. Nina Bradley for teaching me so
much in the neuroanatomy lab, but more so outside the lab: how to teach, engage with
students and navigate life in academia and appreciate it. My committee chair, Kristi
Clark for asking me the right questions to make me understand diffusion, and very
helpful suggestions. I’d also like to thank my committee members Arthur Toga, Daniel
Campbell and Valter Longo for their guidance and ideas that shaped this dissertation.
Derrek Hibar and Neda Jahanshad for incredible intellectual and scientific support,
code, and motivation. Stephen A. Engel, Kathleen Thomas, Bill Kremen, Christine
Fennema-Notestine for mentoring me along the way.
I would like to thank co-authors Jorge Oksenberg, Jacob McCauley, Ashley Beecham,
Katie L. McMahon, Greig I. de Zubicaray, Margaret J. Wright for advice, direction and
very helpful discussions.
I would also like to thank the Multiple Sclerosis Research Group at Oslo University
Hospital, particularly Hanne Flinstad Harbo, for hosting multiple research visits.
None of this would be possible without the support of so many labmates,
roommates, classmates, and friends who’ve been there for me when I need it most.
Talia Nir, David Manheim, Chris Ching, Priya Bhatt, Julio Villalon, Artemis Zavaliangos-
Petropulu, Sarah Madsen, Aggie McMahon, Tuva Hope, Casie and Ann Elyse
Urquidi, Melanie Sweeney, Louise Menendez, Brenton Keller, Rorry Brenner, Katie
Zyuzin, Kirsten Lynch, Ashley Marks.
Last but not least, my parents. The love and support you’ve given me has kept me
going and always will.
I feel so incredibly lucky and cannot express my gratitude enough for all of you.
iii
The outcome of any serious research can only be to make two questions grow where
one question grew before.
—Thorstein Veblen
I would rather live in a world where my life is surrounded by mystery than live in a world
so small that my mind could comprehend it.
—Harry Emerson Fosdick
iv
Table of Contents
Dedication i
Acknowledgements ii
Abstract vii
Chapter 1: Introduction 1
1.1 Polygenic disorders and the post-gwas era 1
1.2 Complex polygenic disorders 1
1.3 Multiple Sclerosis 2
1.3.2. MS Genetics 3
1.4 Coronary Artery Disease 4
1.5 The present study 6
Chapter 2: Genetic pleiotropy between determinants of multiple sclerosis risk
and regional brain volumes.
2.1 Introduction 10
2.2 Methods 10
2.2.1 Samples 10
2.2.2 Data Processing 11
2.2.3 Statistical significance in overlap 11
2.2.4 Detecting pleiotropic variants 12
2.3 Results 13
2.3.1 Pleiotropy 13
2.3.2 Discovery of novel MS susceptibility variants 16
2.4 Discussion 18
2.5 Chapter 2 References 22
Chapter 3: Genetic pleiotropy between determinants of coronary artery disease risk and
regional brain volumes
3.1 Introduction 26
3.2 Methods 26
3.2.1 Description of original association studies 26
3.2.2 Post-processing of genetic data 26
3.2.3 Tests of pleiotropy and concordance 27
3.3 Results
3.3.1 Pleiotropy between globus pallidus,
nucleus accumbens, ICV, and CAD-associated variants 28
3.3.2 Concordance between ICV and CAD-associated variants 28
v
3.4 Discussion 29
3.5 Chapter 3 References 30
Chapter 4: Multiple sclerosis risk variants associated with DTI white-matter integrity
4.1 Introduction 32
4.2 Methods 33
4.2.1 Subjects, genotyping and image acquisition information. 33
4.2.2 Diffusion tensor imaging (DTI) 34
4.2.3 DTI processing 34
4.2.4 Statistical analysis 34
4.2.5 Preliminary Replication cohort 34
4.2.6 The Allen Human Brain Atlas gene expression analysis 35
4.3 Results 35
4.4 Discussion 39
4.5 Chapter 4 References 41
Chapter 5: Multiple Sclerosis polygenic risk scores and white-matter integrity
5.1 Introduction 45
5.2 Methods 46
5.2.1 Subjects and genotyping 46
5.2.2 Polygenic risk scores 46
5.2.3 Risk score association 46
5.2.4 White-matter potholes 46
5.2.5 Gene expression maps 46
5.3 Results 47
5.4 Discussion 50
5.5 Chapter 5 References 51
Chapter 6: Genetic connectivity: correlated genetic control of cortical thickness, brain
volume and white-matter
6.1 Introduction 54
6.2 Methods 55
6.2.2 Subject Information 55
6.2.2 Image Acquisition 55
6.2.3 Image Preprocessing 56
6.2.4 Establishing Zygosity, Genotyping, and Imputation 56
6.2.5 Cross-Twin Cross-Trait Analysis 56
6.2.6 Phenotypic correlations 58
6.2.7 Multiple Comparisons Correction 58
6.3 Results 58
6.4 Discussion 64
6.5 Chapter 6 References 66
vi

Chapter 7: Future directions 68
7.0 Further tests of genetic pleiotropy 68
7.1 Polygenic risk score 68
7.2 Studies with MS patients 68
7.3 Microstructural diffusion MRI 68
7.4 Chapter 7 References 69

vii
Abstract
Virtually all non-Mendelian complex diseases are triggered by the additive effect of
polygenicity and environmental factors. As the past decade has seen an explosion in
GWA studies identifying genetic variants associated with disease risk, the next great
challenge for researchers is to put a functional description on the mechanism driving
these associations. This dissertation describes several imaging genetics studies with
the aim of understanding how genetic risk variants influence multiple sclerosis and
coronary artery disease, two hallmark complex polygenic disorders. Additionally, we
describe several methods for quantifying “genetic connectivity” across traits and
measurements.
Multiple sclerosis is a neurodegenerative demyelinating disease with an unknown
pathogenesis. The disease is highly heterogeneous with a highly complex genetic
architecture. Examining the overlap between genetic loci that affect MS risk and
regional brain structure may reveal mechanisms that influence lesion development in
the brain. Based on recent genomic screens for common variants associated with MS,
we examined pleiotropic overlap between summary statistics from the most recent MS
case-control ImmunoChip study and GWAS summary statistics for seven subcortical
brain volumes (nucleus accumbens, amygdala, caudate nucleus, globus pallidus,
hippocampus, putamen, thalamus), and intracranial volume. We used continuous
inflation analysis (CIA) to test profiles of overlap. We found significant evidence of
overlap between variants in MS risk genes and the volumes of the hippocampus and
thalamus. In follow up analyses of these structures, we used pleiotropy-informed
conditional FDR and identified 24 novel variants associated with both MS risk and
regional brain volumes. This discovery of gene variants associated with both MS risk
and hippocampal and thalamic volumes in the brain may narrow the search for causal
pathways mediating their effect on the human brain.
The mechanism of how single nucleotide polymorphisms (SNPs) confer disease risk is
not completely understood. MRI is utilized clinically and in research as a diagnostic tool
in the identification of the hallmark white-matter lesions. We extracted gene expression
maps of MS risk genes from the Allen Brain Atlas and overlaid them with results from
Rinker et al., 2014 & 2015 where we showed an effect of several MS susceptibility
alleles and polygenic risk score (PRS) on white-matter integrity in healthy young adults,
and replicated these results in healthy aging adults.
The degree to which shared genetic factors influence CAD and brain morphology will
give important insight to the pathogenesis of CAD. Here, using SNP Effect
Concordance Analysis (SECA) we combined summary statistics from the CARDIoGRAM
GWAS with those from Hibar et al. 2015 GWAS of brain volumes, we tested for (1)
global genetic pleiotropy (2) genetic concordance between genetic variants associated
with CAD and brain volume.
viii

MRI and DTI measures of brain volume, cortical thickness and white matter (WM)
integrity are commonly used in imaging genetics studies, but the genetic relationship
between these measures is not well understood. Here we use structural equation
models (SEM) in a twin design to test the genetic correlation between these common
imaging measures. MRI and DTI data from 442 participants (mean age: 23.5 years +/-
2.1 SD; 151 women; 98 MZ pairs, 123 DZ pairs) were analyzed using standardized
ENIGMA protocols. We found significant genetic associations between measure of the
integrity (fractional anisotropy, or FA) of several WM tracts and subcortical volume
ROIs, notably the thalamus and pallidum. Correlation was low between cortical
thickness or volume and DTI measures from the WM. Total cortical surface area was,
however, highly correlated with FA in several WM regions and all of the subcortical
volume regions. These results may be useful for future studies assessing specific
genetic associations, and offer insight into the genetics underlying common imaging
measures.

Finally we describe future plans to extend this work.
CHAPTER ONE
Introduction

1.1 Polygenic disorders and the post-GWAS era

We are now in the “post-gwas era.” Since the first genome-wide association study
(GWAS) was published a decade ago (Thomas et al., 2005), over 2479 unique GWAS
1
studies have been conducted, reporting a total of 297,670 SNP-trait/disease
associations (as of Li et al., 2015). The number of these studies is increasing
exponentially, a phenomenon known as the “genome-wide tide” (Callaway 2017).

Despite this growing trend, GWAS results haven’t exactly lived up to their original
expectations. The advent of the technique was marked by tremendous excitement in
the scientific community. Designed to be answer to the shortcomings of linkage and
candidate gene studies, GWA studies were touted as the “successful new tool for
unlocking the genetic basis of many [...] common causes of human morbidity and
mortality” (Hirschhorn and Daly, 2005). GWA studies search for single nucleotide
polymorphisms (SNPs) that are associated with common complex diseases, and it was
thought that this would finally be the key to unlock polygenic disease. Puzzlingly, GWAS
has not yet solved a single polygenic disease and we are left looking for explanations
for the missing heritability— the mysterious gap between calculated heritability and that
which should be explained by GWAS.

Researchers are currently changing their views of GWAS from being to able to fully
explain disease, to seeing it as a preliminary result that requires further investigation.
Proposed tools include testing GWAS results for epistasis, pathway analysis of GWAS
results and alternative ways to prioritize SNPs unveiled by GWAS (Cantor, Lange, and
Sinsheimer, 2010).

1.2 Complex polygenic disorders

While some disorders are caused by a mutation in a single gene, such as cystic fibrosis
or sickle cell disease— most common disease in humans have multifactorial. Such
disorders are known as complex diseases and virtually all of them are triggered by the
additive effect of many genes (polygenicity) and environmental factors.

Two particular complex polygenic disorders are the focus of this dissertation: Multiple
Sclerosis and Coronary artery disease.

1
Though, it wouldn’t be until 2007 that the detailed method would be described in the Handbook
of Statistical Genetics (Morris et al., 2007).
1

1.3 Multiple Sclerosis

Multiple sclerosis (MS) is a neurodegenerative inflammatory disease and leading cause
of neurologic dysfunction in young adults, currently with no known cause or cure. A
complex interplay of genetic and environmental risk factors is thought to promote the
illness (Hassan Smith, and Douglas, 2011). Epidemiological studies point to several
candidate environmental risk factors: living in more northern latitudes, vitamin D
deficits, toxin or possibly virus exposure (Ascherio, 2013) whereas twin and family
studies show that genetic factors are a major determinant of risk (Baranzini, 2011).

The three hallmarks of the disease are the 1). The destruction of myelin in the central
nervous system (CNS), 2) the formation of lesions or plaques in the CNS, and 3)
inflammation. The symptoms of MS are broad and can be explained by slowed or
blocked neural signals due to breakdown of myelin affecting action potentials or
complete degeneration of neurons. Inflammation is the primary cause of the
neurodegeneration (Loma and Heyman, 2011). It is thought to be caused by both innate
and adaptive immune responses, coupled by blood brain barrier breakdown.

Environmental factors associated with MS have long been documented.
Epidemiological findings show that rates of MS are much higher in people who live the
furthest from the equator, in both hemispheres (Ascherio, 2013). Vitamin D deficiency
due to lack of sunlight exposure is the prevailing hypothesized explanation for this
phenomenon. Researchers have thoroughly tested this with a satellite data, showing
that areas with low levels of UVB were 20 times more likely to have higher prevalence of
MS. Migration studies have also corroborated this hypothesis, finding that moving from
a high to low risk area early in life reduces MS risk (Ascherio, et al., 2014). Conversely,
various studies have shown sunlight exposure early in life to be protective against MS
(Pierrot-Deseilligny, 2009). For this reason, Vitamin D therapy has been studied in MS
treatment, with limited success. (Salzer, Biström and Sundström, 2014).

Other environmental factors include smoking, toxin exposure and possibly infectious
agent exposure; particularly human herpesvirus, Epstein Barr virus and mycoplasma
pneumoniae (Pakpoor and Ramagopalan, 2013). Stress has also been implicated,
although results are mixed. T2 gadolinium enhancing lesion count was correlated with
stress exposure in one study (Lovera and Reza, 2013). Recently, oxidative stress has
been explored as risk factor in MS, and related diet and genetic factors have been
proposed as protective (Ong, et al., 2015; Liang, et al., 2015).

MS is a very mysterious disease. Despite the fact that CD4(+) effector T cells have
repeatedly been implicated in MS lesions, treatments targeting them are not effective.
(Kasper and Shoemaker, 2010). Pregnancy significantly reduces the likelihood of acute
MS attacks, though researchers are still unsure why— hormone changes influencing
2

underlying immune responses are likely (Miller et al., 2014). There is also the problem of
the “clinical MR paradox,” where despite being required for a diagnosis, there is often
no correlation between imaged lesions and disease status. (Kacar, et al., 2011). These
are only a few examples highlighting the enigmatic nature of the disease.

1.3.2 MS Genetics

There has been extensive research on the genetic architecture of MS: microarray
studies, candidate gene studies, linkage studies, GWA studies being the primary
methods. The disease has been found to cluster in families; MS increases with shared
genetics, however, it is not a simple Mendelian disorder. Monozygotic twin concordance
rates have been shown to be at ~25% 30%, while twins and full siblings are at 25%.
The monozygotic concordance rate of 30% indicates a complex genetic component,
with substantial environmental influence. It also much more prevalent in individuals with
European ancestry, although it does occur in other populations as well. (McElroy and
Oksenberg, 2010)

Numerous genetic studies have shown the Human Leukocyte Antigen (HLA) genes of
the Major Histocompatibility Complex (MHC) provide the strongest and most reliable
genetic risk signal for MS (Moutsianas et al., 2015; Patsopoulos et al., 2013). HLA
proteins are associated with autoimmune response and self recognition of antigens. A
wide array of loci in this region have been associated with MS through various genetic
techniques. There is not a clear consensus as to which genes in this region confer the
greatest impact on MS etiology, or their exact mechanism. It is possible that several are
involved in an epistatic pathway.

Prior to the advent of modern GWAS methods, there was little success finding
legitimate MS associated loci outside of the HLA. The only other region with a modicum
of effect and replication is from the interleukin receptor 7 gene (IL7R), which is also
related to autoimmune function. Variants on this gene and also IL2RA were the first to
reach genome wide significance in early GWA studies on MS. Since the first two
relatively small GWA studies, there have been 14 more. The largest to published date
was lead by a combined effort by the International MS Genetics Consortium (IMSGC)
and the Wellcome Trust Case Control Consortium (WTCC2). This massive GWAS of
9,772 cases confirmed 23 previously reported variants and discovered 34 new
associated variants (five of which just missed the genome wide significance threshold,
but were later verified in a followup study by the IMSGC).

Preliminary results from the IMSGC’s latest and most comprehensive genetic study of
MS (currently posted to bioRxiv) can explain ~39% of the genetic predisposition for MS,
finding significant enrichment of MS susceptibility genes in cells of both the innate and
adaptive immune system (Patsopoulos et al., 2017). That study expands upon findings
unveiled by the ImmunoChip, a custom genotyping chip designed to improve signal in
3

immunogenetic studies by focusing the coverage towards deep replication of loci
implicated by previous GWAS of 12 autoimmune diseases (IMSGC, 2017). With the
Immunochip, Beecham et al., reported the list of known risk SNPs to account for an
estimated 28% of the genetic risk for MS, and revealed the single-nucleotide


polymorphism (SNP) rs12087340 as the most significant novel association for MS,
located near gene BCL10— a potential locus for intervention associated with
inflammation, immunity, and apoptosis. Another novel SNP discovered in that
experiment was exonic missense variant rs2288904, which is known to be an upstream
factor in T-cell expression and various aspects of inflammation.

One promising thing about the GWAS findings is that the 57 SNPs show strong putative
functional coherence. 21 of the 57 have been previously associated with other
autoimmune diseases or are in linkage disequilibrium with those that are.

GWAS evidence has strongly influenced theories of autoimmune dysfunction in MS.
The strong association with immune function led to the collaborative development of
the Immunochip— a SNP chip focused on fine mapping of loci related to immune
system function; it is not representative of the entire genome but provides increased
resolution in mapping immune related loci. A massive Immunochip analysis with 14,802
cases and 26,703 controls yielded 45 new loci, many within the MHC or associated with
immune function. Despite these promising results, MS genetic researchers lament much
of the missing heritability is likely found in regions that fall short of the “essentially
arbitrary, p<5 × 10−8 threshold” (Sawcer, et al., 2014). It also is likely explained by
polygenicity and perhaps structural genomic variation.

1.4 Coronary Artery Disease

Coronary artery disease (CAD) is the most common cause of death worldwide (Go et
al., 2013, WHO 2012). Recent efforts in epidemiology and quantitative genetics have
shown that genetic predisposition is equivalent to traditional and environmental factors
for disease risk (McPherson and Tybjærg-Hansen 2016).

The disease has been estimated to be between 40-60% heritable (Vinkhuyzen et al.),
and is considered to be a classic complex polygenic disorder where a number of
common single-nucleotide polymorphisms (SNPs) contribute to disease susceptibility
(Roberts 2014). The Coronary ARtery DIsease Genome wide Replication and
Meta-analysis (CARDIoGRAM) consortium published a genome wide association study
(GWAS) of CAD comprising of 63,746 cases and 130,681 controls— the largest genetic
assessment of CAD to date (Schunkert 2013). That study increased the list of known
CAD risk SNPs to a total of 45, and a recent meta-analysis effort reports that number to
be 48, explaining nearly 20% of the heritability (Nikpay et al 2015).

4

There is a well-established link between CAD and cerebrovascular disease including
silent brain injury, vascular dementia, and increased stroke risk (Gąsecki 2013). CAD
and several neurological disorders share overlapping environmental risk factors
(Kovacic et al 2012, Muqtadar 2012). Neuroimaging studies have found associations
between cognitive dysfunction and CAD, suggesting a cerebrovascular etiology (Vogels
et al 2007a & 2007b, Eggermont et al, 2012). A number of recent studies report reduced
brain volume and white-matter integrity in CAD patients when compared to healthy
controls. Hypertension—an archetypal CAD risk factor—has been associated with
differences in white-matter microstructure detectable as early as middle age (McEvoy et
al., 2015).

Both coronary artery disease susceptibility and brain volume may be influenced by
confounding risk factors or masked by survival bias. There is a paucity of literature
establishing the mechanistic link between the two (Mahmood et al., 2013).

The CARDIoGRAM study found several CAD variants that are associated with various
other diseases and traits. It is not clear whether these and other putative pleiotropic
variants represent genuine multiple effects, concerted pathogenic mechanisms, or
simply chromosomal regions with a greater probability of variation. Additionally, as CAD
develops over the lifespan, antagonistic pleiotropy likely explains why Darwinian
selection has not eliminated the disease— the same genes that contribute to CAD
susceptibility partially promote reproduction (Byars et al. 2017).


5

1.5 Chapter 1 References

Ascherio A. Environmental factors in multiple sclerosis. Expert Rev Neurother. 2013
Dec;13(12 Suppl):39. doi: 10.1586/14737175.2013.865866. Review. PubMed PMID:
24289836.

Ascherio A, Munger KL, White R, Köchert K, Simon KC, Polman CH, Freedman MS,
Hartung HP , Miller DH, Montalbán X, Edan G, Barkhof F , Pleimes D, Radü EW,
Sandbrink R, Kappos L, Pohl C. Vitamin D as an early predictor of multiple
sclerosis activity and progression. JAMA Neurol. 2014 Mar;71(3):30614. doi:
10.1001/jamaneurol.2013.5993. PubMed PMID: 24445558; PubMed Central PMCID:
PMC4000029.

Balding DJ, Bishop M, Cannings C, editors. Handbook of statistical genetics. John
Wiley & Sons; 2008 Jun 10.

Baranzini SE. Revealing the genetic basis of multiple sclerosis: are we there yet? Curr
Opin Genet Dev. 2011 Jun;21(3):31724. doi: 10.1016/j.gde.2010.12.006. Epub 2011 Jan
17. Review. PubMed PMID: 21247752; PubMed Central PMCID: PMC3105160.

Byars SG, Huang QQ, Gray LA, Bakshi A, Ripatti S, Abraham G, Stearns SC, Inouye M.
Genetic loci associated with coronary artery disease harbor evidence of selection and
antagonistic pleiotropy. PLoS genetics. 2017 Jun 22;13(6):e1006328.

Callaway E. New concerns raised over value of genome-wide disease studies. Nature
News. 2017 Jun 22;546(7659):463.

Cantor RM, Lange K, Sinsheimer JS. Prioritizing GWAS results: A review of statistical
methods and recommendations for their application. Am J Hum Genet. 2010
Jan;86(1):622. doi: 10.1016/j.ajhg.2009.11.017. Review. PubMed PMID: 20074509;
PubMed Central PMCID: PMC2801749.

Eggermont LHP , de Boer K, Muller M, et al Cardiac disease and cognitive impairment: a
systematic review Heart 2012;98:1334-1340.

Gąsecki D, Kwarciany M, Nyka W, Narkiewicz K. Hypertension, brain damage and
cognitive decline. Current hypertension reports. 2013 Dec 1;15(6):547-58.

Go AS, Mozaffarian D, Roger VL, et al. Heart disease and stroke statistics—2013
update: a report from the American Heart Association. Circulation. 2013;127:e6–e245.

Hassan-Smith G, Douglas MR. Epidemiology and diagnosis of multiple sclerosis. British
journal of hospital medicine (London, England: 2005). 2011 Oct;72(10):M146-51.
6
Hirschhorn JN, Daly MJ. Genomewide association studies for common diseases and
complex traits. Nat Rev Genet. 2005 Feb;6(2):95108. Review. PubMed PMID:
15716906.
Hirschhorn JN, Daly MJ. Genome-wide association studies for common diseases and
complex traits. Nature Reviews Genetics. 2005 Feb 1;6(2):95-108.
International Multiple Sclerosis Genetics Consortium. Analysis of immune-related loci
identifies 48 new susceptibility variants for multiple sclerosis. Nature genetics . 2013
Nov 1;45(11):1353-60.
Kacar K, Rocca MA, Copetti M, Sala S, Mesaroš Š, Opinćal TS, Caputo D, Absinta M,
Drulović J, Kostić VS, Comi G. Overcoming the clinical–MR imaging paradox of multiple
sclerosis: MR imaging data assessed with a random forest approach. American Journal
of Neuroradiology. 2011 Dec 1;32(11):2098-102.
Kasper LH, Shoemaker J. Multiple sclerosis immunology The healthy immune system vs
the MS immune system. Neurology. 2010 Jan 5;74(1 Supplement 1):S2-8.
Kovacic JC, Mehran R, Karajgikar R, Baber U, Suleman J, Kim MC, Krishnan P , Dangas
G, Sharma SK, Kini A. Female gender and mortality after percutaneous coronary
intervention: results from a large registry. Catheterization and Cardiovascular
Interventions. 2012 Oct 1;80(4):514-21.
Li MJ, Liu Z, Wang P , Wong MP , Nelson MR, Kocher JP , Yeager M, Sham PC, Chanock
SJ, Xia Z, Wang J. GWASdb v2: an update database for human genetic variants
identified by genome-wide association studies. Nucleic acids research. 2015 Nov
28;44(D1):D869-76.
Liang P , Le W. Role of autophagy in the pathogenesis of multiple sclerosis.
Neuroscience bulletin. 2015 Aug 1;31(4):435-44.
Loma I, Heyman R. Multiple sclerosis: pathogenesis and treatment. Current
neuropharmacology. 2011 Sep 1;9(3):409-16.
Lovera J, Reza T. Stress in multiple sclerosis: review of new developments and future
directions. Current neurology and neuroscience reports. 2013 Nov 1;13(11):398.
Mahmood S. (2009) Pathophysiology of Coronary Artery Disease. In: Movahed A.,
Gnanasegaran G., Buscombe J., Hall M. (eds) Integrating Cardiology for Nuclear
Medicine Physicians. Springer, Berlin, Heidelberg
7
McElroy JP , Oksenberg JR. Multiple sclerosis genetics 2010. Neurologic clinics. 2011
May 31;29(2):219-31.
McPherson R, Tybjærg-Hansen A. Genetics of Coronary Artery Disease. Circulation
research. 2016 Feb;118(4):564-78.
Miller DH, Fazekas F , Montalban X, Reingold SC, Trojano M. Pregnancy, sex and
hormonal factors in multiple sclerosis. Multiple Sclerosis Journal. 2014
Apr;20(5):527-36.
Morris AP , Cardon LR, Balding DJ, Bishop M, Cannings C. Handbook of statistical
genetics.
Moutsianas L, Jostins L, Beecham AH, Dilthey AT, Xifara DK, Ban M, Shah TS,
Patsopoulos NA, Alfredsson L, Anderson CA, Attfield KE, Baranzini SE, Barrett J, Binder
TM, Booth D, Buck D, Celius EG, Cotsapas C, D'Alfonso S, Dendrou CA, Donnelly P ,
Dubois B, Fontaine B, Lar Fugger L, Goris A, Gourraud PA, Graetz C, Hemmer B, Hillert
J; International IBD Genetics Consortium (IIBDGC)., Kockum I, Leslie S, Lill CM,
Martinelli-Boneschi F , Oksenberg JR, Olsson T, Oturai A, Saarela J, Søndergaard HB,
Spurkland A, Taylor B, Winkelmann J, Zipp F , Haines JL, Pericak-Vance MA, Spencer
CC, Stewart G, Hafler DA, Ivinson AJ, Harbo HF , Hauser SL, De Jager PL, Compston A,
McCauley JL, Sawcer S, McVean G; International Multiple Sclerosis Genetics
Consortium. Class II HLA interactions modulate genetic risk for multiple sclerosis. Nat
Genet. 2015 Oct;47(10):1107-13.
Muqtadar H, Testai FD, Gorelick PB. The dementia of cardiac disease. Current
cardiology reports. 2012 Dec 1;14(6):732-40.
Nikpay, M., Goel, A., Won, H.-H., Hall, L. M., Willenborg, C., Kanoni, S., ... Farrall, M.
(2015). A comprehensive 1000 Genomes-based genome-wide association
meta-analysis of coronary artery disease. Nature Genetics , 47 (10), 1121–1130.
Ong WY, Farooqui T, Koh HL, Farooqui AA, Ling EA. Protective effects of ginseng on
neurological disorders. Frontiers in aging neuroscience. 2015;7.
Pakpoor J, Ramagopalan SV. Epstein-Barr virus is a necessary causative agent in the
pathogenesis of multiple sclerosis: Yes. Multiple Sclerosis Journal. 2013 Nov
1;19(13):1690-1.
Patsopoulos NA, Barcellos LF , Hintzen RQ, Schaefer C, Van Duijn CM, Noble JA, Raj T,
Gourraud PA, Stranger BE, Oksenberg J, Olsson T. Fine-mapping the genetic
association of the major histocompatibility complex in multiple sclerosis: HLA and
non-HLA effects. PLoS Genet. 2013 Nov 21;9(11):e1003926.
8

Patsopoulos NA, Baranzini SE, Santaniello A, Shoostari P , Cotsapas C, Wong G,
Beecham AH, James T, Replogle J, Vlachos I, McCabe C. The Multiple Sclerosis
Genomic Map: Role of peripheral immune cells and resident microglia in susceptibility.
bioRxiv. 2017 Jan 1:14393

Pierrot-Deseilligny C. Clinical implications of a possible role of vitamin D in multiple
sclerosis. Journal of neurology. 2009 Sep 1;256(9):1468-79.

Roberts, R. (2014). Genetics of Coronary Artery Disease: An Update. Methodist
DeBakey Cardiovascular Journal , 10 (1), 7–12.

Salzer J, Biström M, Sundström P . Vitamin D and multiple sclerosis: where do we go
from here?. Expert review of neurotherapeutics. 2014 Jan 1;14(1):9-18.

Sawcer S, Franklin RJ, Ban M. Multiple sclerosis genetics. The Lancet Neurology. 2014
Jul 31;13(7):700-9.

Schunkert, H. , et al. Large-scale association analysis identifies 13 new susceptibility
loci for coronary artery disease. Nat Genet 43, 333-338 (2011).

The European health report 2012: charting the way to well-being. Copenhagen,
Denmark: WHO Regional Office for Europe; 2012.

Thomas DC, Haile RW, Duggan D. Recent developments in genomewide association
scans: a workshop summary and review. The American Journal of Human Genetics.
2005 Sep 30;77(3):337-45.

Vinkhuyzen AA, Wray NR, Yang J, Goddard ME, Visscher PM. Estimation and partition
of heritability in human populations using whole-genome analysis methods. Annual
review of genetics. 2013 Nov 23;47:75-95.

Vogels R, L, C, Oosterman J, M, van Harten B, Gouw A, A, Schroeder-Tanka J, M,
Scheltens P , van der Flier W, M, Weinstein H, C, Neuroimaging and Correlates of
Cognitive Function among Patients with Heart Failure. Dement Geriatr Cogn Disord
2007;24:418-423

Vogels RL, Scheltens P , Schroeder-Tanka JM, Weinstein HC: Cognitive impairment in
heart failure: a systematic review of the literature. Eur J Heart Fail 2007;9:440–449.


9
CHAPTER TWO
Genetic pleiotropy between determinants of
multiple sclerosis risk and regional brain volumes.
2.1 Introduction
See Chapter 1 for introduction to MS genetics.
In the last decade, large genome-wide association studies (GWAS) and subsequent
follow-up efforts have identified over 200 common genetic variants outside of the MHC
with differing frequencies in MS patients versus controls, pointing to genetic drivers of
MS susceptibility (Patsopoulos et al., 2013 and 2017). However, the specific genetic
variants promoting the disease and how they affect the healthy and degenerating MS
central nervous system remain largely unknown.
Brain atrophy, particularly volume loss, is a hallmark of MS, usually occurring early in
the disease (IMSGC, 2013). Compared to age-matched healthy controls, structural


brain atrophy in most MS patients advances more rapidly (Miller et al., 2002; Bermel et
al., 2006). Specifically, abnormalities in the thalamus are of clinical relevance, as it is a
keystone hub connecting fibers that may be affected by the disease (Liu et al., 2015).
These structural changes are relevant in regards to MS symptom presentation, but their
causes are unknown (Bermel et al., 2006).
Recently, Hibar and colleagues screened the genomes and MRI scans of 30,717
subjects, revealing genetic loci associated with numerous subcortical volumes,
including brain regions implicated in MS (Hibar et al., 2015). In the present study, we
used a novel statistical technique that combines summary statistics from an
ImmunoChip-based genetic analysis, the largest published genetic study in MS to date
(IMSGC 2013), with summary statistics from the subcortical brain structure GWAS in
ENIGMA (Hibar et al., 2015) to (1) test for genetic overlap, or pleiotropy, between
genetic modulators of MS risk and brain structure and (2) to identify additional
susceptibility variants for MS based on their impact on specific brain structures.
2.2 Methods
2.2.1 Samples
We obtained summary statistics available from the largest available published genetic
analysis of MS to date (IMSGC, 2013) comprising 14,498 patients diagnosed with MS
and 24,091 healthy controls. In addition, we obtained summary statistics available from
the largest GWAS of intracranial volume (ICV) and seven subcortical volumes to date
from the ENIGMA Consortium (Hibar et al., 2015). Participants in both studies were
primarily of European descent.
10
2.2.2 Data Processing
The original set of summary statistics from the ENIGMA2 GWAS consists of 8,398,366
SNPs available across all eight brain structures. The original MS summary statistics
consist of 161,311 SNPs. When merging the SNP datasets across MS and brain
structures, a total of 107,632 SNPs were present in all datasets. In order to make
genomic comparisons between MS and brain structure datasets we first obtained a set
of independent SNPs representing distinct linkage disequilibrium (LD) blocks across the
genome. We performed an LD clumping procedure on each of the ENIGMA2 brain
volume GWAS using the subjects of European ancestry from the 1000 Genomes to
calculate LD block coordinates in PLINK (Purcell et al., 2007) (with the following
parameters: physical distance threshold, 500 kb; LD threshold for clumping, r
2
=0.2; and
a P-value cutoff of 1). This resulted in a list independent, overlapping SNPs, which we
carried forward for analysis. This was done for each brain structure, ranging from
24,271 to 25,085 SNPs per structure, respectively.
2.2.3 Statistical significance in overlap
We used a threshold-free inflation analysis, termed continuous inflation analysis (CIA),
(Hibar et al., 2015) to examine the statistical significance in overlap between the two
datasets without having to impose pre-defined P-value-based cutoffs in the statistics of
either dataset, as is the case with polygenic scoring. This avoids missing signal from
genomic areas of lower significance, which should not be overlooked when polygenicity
is high. CIA is a recently developed algorithm designed to assess global evidence of
enrichment between any two traits, and extends the pleiotropy test from Andreassen et
al., (2013) We declare a comparison significant if the global test of overlap exceeds a
Bonferroni-corrected significance threshold accounting for eight comparisons (Pthresh
= (0.05/8) = 6.25x10
-3
)
CIA is a threshold-free method to assess global enrichment in genetic loci based on
their evidence of being associated with two different traits. First, we perform clumping
on each brain volume GWAS using the European subjects from the 1000 Genomes
reference panel to calculate LD block coordinates. Clumping was run in PLINK with the
following options: (—clump —clump-kb 500 —clump-p1 1 —clump-p2 1 —clump-r2
0.2). During the clumping process we identify a single, independent SNP from each LD
block and then merge the MS Immunochip result with each brain volume GWAS result
such that only the independent SNPs from both datasets make into the final merged
dataset (here, the analysis is performed 8 times, one for each brain volume GWAS:
nucleus accumbens, amygdala, caudate nucleus, globus pallidus, hippocampus,
putamen, thalamus, and intracranial volume). Next, the merged datasets are sorted in
descending order based on the P-value for a given SNP in each respective brain volume
GWAS. We then have a matching list of SNPs for the MS Immunochip and each brain
11
volume GWAS, ranked by SNP P-value in the brain volume GWAS. In order to calculate
the enrichment, we iterate through the P-value-ordered SNP list at a given step size, n =
10, with each step moving down the list by n SNPs.
2.2.4 Detecting pleiotropic variants
For brain volumes that showed genomic overlap with MS risk—the hippocampus and
thalamus—we performed a test of enrichment at the peak cutoff in the CIA plot. The
peak cutoff in each plot is based on the peak enrichment value in the MS dataset when
conditioning on variants underlying brain volume. In this way, we select a prioritized
subset of SNPs from the original complete MS ImmunoChip dataset that show higher
evidence of association with regional brain volumes.
The level of enrichment in MS conditioned on a subset of SNPs prioritized by the effect
on a given brain structure is indicated with a lambda inflation factor, which is computed
at each step using all of the P-values from the MS Immunochip below the SNP at the
current step level in the list. This procedure iterates until the step size n is greater than
the total number of SNPs left in the list.
We determine a significance level for each of the lambda inflation factors at each step
using a permutation-based approach. By running m = 10,000 permutations, we permute
the whole list of MS SNPs and then begin the CIA process again; in this way, each step
gets a lambda inflation factor which is stored for each step and permutation. A
one-sided nonparametric P-value for each permuted lambda inflation factor is
calculated:
((number of permuted lambda values > permuted lambda at a given step and
permutation) + 1)/(m + 1)
This process terminates when a P-value is generated for each step and permuted
distribution. Within each permutation, we combine P-values across each step using
Fisher’s P-value combining procedure. This gives us a chi-squared statistic for each
permutation, which is then combined across permutations to generate the expected
null distribution given our data. We then also calculate a one-sided, nonparametric
P-value for our observed, non-permuted data, and combine the P-values using Fisher’s
combining procedure. The global enrichment P-value is then calculated:
((number of null chi-squared statistics > observed chi-squared statistic) + 1)/(m + 1)
Global enrichment and pleiotropy is considered significant if the global P-value <
6.25x10
-3
. The set of SNPs in the most enriched cut off were extracted for the MS
Immunochip dataset along with the corresponding P-value representing each SNP’s
12
association with MS risk. The false discovery rate method, FDR (Benjamini and
Hochberg, 1995), is applied to SNPs from the most significant CIA cutoff in the MS
Immunochip P-values (given by the cutoff in the CIA that gives the largest enrichment
value lambda). We then applied the FDR procedure to the full set of MS Immunochip
SNPs used in the CIA in order to show which variants would have been declared
significant without enrichment subsets. Again, SNPs were considered significant at a
5% false discovery rate (q < 0.05). Significant SNPs from this conditional FDR set that
were not significant in the full set, are considered to be detected with greater power due
to conditioning on variants influencing brain volume.
2.3 Results
2.3.1 Pleiotropy
We discovered significant evidence of pleiotropy between genomic variants associated
with MS risk and variants that modulate the volumes of two subcortical brain structures:
the hippocampus (P < 1.0e-5) and thalamus (P < 0.001). Variants affecting the putamen
(P = 0. 0.00299) and Caudate (P = 0. 0.00499) showed trends for pleiotropic overlap
with MS.
13
Figure 2.1 CIA analysis: lambda inflation factor in MS vs Hippocampal GWAS. The x-axis shows SNPs from the
hippocampal volume GWAS ranked by –log10(P-value). y-Axis shows lambda inflation factor in MS. Lower blue dotted
line represents the expected value, while the upper blue line represents the 95
th
percentile. Red points are the LIF at each
iteration. Points outside of the upper blue dotted line represent SNPs considered to be significantly enriched. Asterisk
denotes the peak cutoff.
14
Figure 2.2 CIA analysis: lambda inflation factor in MS vs Thalamic GWAS. The x-axis shows SNPs from the thalamic
volume GWAS ranked by –log10(P-value). y-Axis shows lambda inflation factor (LIF) in MS. Lower blue dotted line
represents the expected value, while the upper blue line represents the 95
th
percentile. Red points are the LIF at each
iteration. Points outside of the upper blue dotted line represent SNPs considered to be significantly enriched. Asterisk
denotes the peak cutoff.
15
We did not find evidence for significant pleiotropic overlap between MS and other
structures tested: accumbens (P = 0.1), amygdala (P = 0.55), and pallidum (P = 0.02).
Variants mediating ICV also did not relate to MS risk in these analyses (P= 0.01).
2.3.2 Discovery of novel MS susceptibility variants
In follow up analysis of the hippocampus and thalamus, we used pleiotropy-informed
conditional FDR to discover 26 significant (q < 0.05) MS-associated candidate variants,
24 of which were undetected in previous large GWAS studies in MS (IMSGC 2007,
2011; IMSGC, 2013). Nineteen significant SNPs were identified when conditioning MS
GWAS results on hippocampal volume and the remaining 7 while conditioning on
thalamic volume (two SNPs were significant in both). Individual SNPs are listed in Tables
1 and 2. In order to determine whether the MS-associated variants from the conditional
analysis represent novel associations or are tagging previously known associations we
estimated the LD between the genome-wide significant SNPs from Beecham et al.,
2013 and the IMSGC 2011, with the SNPs identified in the conditional FDR analysis.
The linkage disequilibrium between SNP pairs was calculated using the 1000 Genome
Phase 1 European population (Sivakumaran et al., 2011). We found that one of the
SNPs (rs10931481) has R
2
>0.2 with previously identified SNPs from large GWAS
studies in MS (rs9967792). We also examined whether any of the significant SNPs from
the conditional FDR analysis were correlated with any proxy SNPs of the SNPs
identified in previous GWAS studies in MS. We found one additional variant (rs4915470)
was linked to a previously identified variant at R
2
> 0.2 (rs55838263). These two
aforementioned SNPs likely represent signal from previously reported loci and are not
considered novel candidate variants and are not included in Tables 1 and 2. Novel SNPs
within 1MB of previously reported MS variants are reported in Tables 1 and 2 with
corresponding LD values. LD values were modest for all associations (max = 0.0374 R
2
).
16

Table 2.1: Hippocampal SNPs

Table 2.2: Thalamic SNPs

Using the same approach, we also tested for overlap with results from the currently
unpublished IMSGC bioRxiv report (Patsopoulos et al., 2017). We found that 16 of the
SNPs revealed in our conditional FDR analysis were within 1 MB of novel variants
discovered independently by the IMSGC in the latest report, however only 2 are in LD
with reported SNPs (with modest R
2
values: .1959 and .1792). These results are
reported in Table 3.


17
Table 2.3: Newly discovered variants here and in Patsopoulos et al., IMSGC bioRxiv preprint.
2.4 Discussion
We found clear genetic overlap between MS-associated genetic variants and those that
modulate subcortical brain volume in the hippocampus and thalamus. This finding
builds upon research suggesting that MS is highly polygenic with a multitude of
pleiotropic variants contributing to the disease (Sivakumaran et al., 2011; Lauc et al.,
2013; Andreassen et al., 2015).
Using the conditional FDR procedure in CIA, we found suggestive evidence that a
number (24) of previously unidentified variants are associated with MS risk by selecting
sets of SNPs with greater associations with brain structure. By pooling summary
statistics from two independent, yet functionally related genetic association studies
(one full GWAS, the other ImmunoChip) we can boost power to detect these variants,
which were not genome-wide significant or detectable using less conservative methods
like the false discovery rate (FDR) in the original independent samples (see Tables 1 and
2). This shows the feasibility of combining joint information from different genomic
screens to boost power to detect novel variants using CIA. This offers a cost-effective
18
way to derive additional information from current GWAS findings of quantitative brain
traits and neurological disorders with clear brain related pathogenesis.
The power to detect novel variants comes from conditioning on brain morphology
related variants, which are ranked in a subset based on the peak inflation value. For
each test, the statistical rankings will differ for each brain region. We are not able to
clinically predict that MS lesions are more likely to appear in one brain region than
another, however, tissue expression patterns in known MS risk genes differ across brain
regions. There are emerging clues that the spatial distribution of MS lesions may be
driven by gene expression as part of an inflammatory immune system response
(Gourraud et al., 2013). MS risk variants identified here should be targets for future gene
enrichment and expression studies.
Variants affecting volume in the hippocampus and thalamus showed pleiotropy with MS
risk while several others failed to reach significance. Grey matter damage, including in
the hippocampus and thalamus, is clinically relevant in MS and possibly occurs
independently of white-matter demyelination (Geurts and Barkhof 2008). The
hippocampus additionally has been shown to be specifically vulnerable (Sicotte et al.,
2008).
An additional simple explanation as to why the hippocampus and thalamus provided
our strongest results could be that these particular regions provide strong signals of
genetic variance. Eyler et al., placed hippocampal volume as the strongest driver of one
of four independent genetic factors of brain morphology, while the basal
ganglia/thalamus was its own factor entirely (Eyler et al., 2011).
Concurrent to the present study, the IMSGC independently performed the largest and
most comprehensive MS genetic association study as a follow-up to the ImmunoChip
(Patsopoulos et al., 2017). While currently unpublished, preliminary results from that
study are posted to Biorxiv, an online preprint server. The report identifies 90 novel MS
autosomal effects outside the MHC, putting the total of known risk variants to 200
SNPs, an increase from the previous 110 at the time of the ImmunoChip publication
(IMSGC 2013). Two variants identified here are in modest LD with newly identified
variants from that study, and 16 of ours are within 1 MB. It is likely that many of these
SNPs are tagging the same signal. This verifies conditional FDR using CIA as a
technique and provides preliminary replication of the IMSGC’s latest findings.
The variant rs9937837 on gene ITGAM reached conditional FDR significance for both
thalamic and hippocampal volumes. ITGAM encodes leukocyte-specific integrin
membrane protein antigens, which function as macrophage receptors (Mac-1). Mac-1
knockout mice show altered inflammatory function in the context of vascular disease
(Simon et al. 2000). ITGAM has previously been speculated to be involved in MS
pathology, as it was found to be included in an immune related pathway significantly
19
related to MS (Baranzini 2011; Anaya et al., 2012). It has also been associated with the
autoimmune disease systemic lupus erythematosus. rs12739786, within 1.6kb 3' of
SCARNA16 (Small Cajal Body-Specific RNA 16), was also significant for both structures
and is in modest LD with several variants near or in SCARNA16, which encodes a small
non-coding RNA gene on chromosome 17 and regulates translation through
interactions with ribosomal RNA subunits.
Novel variants were detected in several other notable genes: DDC (rs11575543, which
is in perfect LD with missense SNP rs11575542), APLP1 (rs230261, in perfect LD with
SNPs on KIRREL2 and NFKBID), and DGKQ (rs4690169). DDC codes for an important
enzyme for decarboxylation of neurotransmitter precursors L-DOPA and 5-HTP into
dopamine and serotonin, respectively. APLP1 is important for synaptic development,
and may play an upstream role in neuronal apoptosis (Wasco et al., 1993). Deficiencies
in DGKQ eye-specific homolog DAGK cause retinal degeneration in Drosophila (Endele
et al., 1996). Two other missense SNPs worth reporting were also discovered:
rs7100382 in C10orf112 and rs61753197 in NCAPD2. C10orf112 is highly expressed in
the human brain and is typically associated with Alzheimer’s disease risk as it encodes
a protein known to interact with APOE4, the major known risk haplotype for late-onset
AD (Zubenko and Hughes, 2009). NCAPD2 is essential for mitotic chromosome
condensation and is expressed in all human tissue types (Schmiesing et al., 2000;
Nagase et al., 1995).
A full list of SNPs may be found in Tables 1 and 2. It is difficult to speculate how these
individual SNPs may functionally contribute to disease risk, but in the future—as the
pathogenesis of MS and other immune system disorders becomes clearer—we may
discover how variants work together, perhaps reaching a threshold that leads to
autoimmune dysfunction or other hypothesized etiologies (Lehmann et al., 2015).
Further investigation of the function of these variants, and others associated with MS
risk, are required.
One limitation of the study comes from inherent limitations of the ImmunoChip. While it
contributed to the most comprehensive published list of MS loci to date, it still relies on
the prior GWA studies from which it was modeled (most of the loci from previous MS
GWA studies were included in the array), and does not cover the whole genome. In fact,
the ImmunoChip contains immune related loci across 12 autoimmune diseases, so we
are more likely to focus on immune related loci. Variants left undetected due to lack of
coverage in prior studies are not evaluated here, and their effects, if present, remain
undetected despite the potential boost in power (rare variants may still be undetected
given the sample size). The focus of the ImmunoChip also precluded us from employing
other methods for testing genetic overlap— such as LD score regression which requires
greater coverage in the genome (Bulik-Sullivan et al., 2015) or SNP effect concordance
analysis, which would need a larger list of SNPs to achieve satisfactory power (Nyholt et
al., 2014).
20
Another limitation of this study is that the genetic profiles are based on whites of
European descent and do not fully represent other ethnic groups. A recent ImmunoChip
study of MS risk in African-Americans suggested significant overlap between African
American and European risk SNPs for MS, but the degree of overlap is not completely
clear (Isobe et al., 2015; Zhuange et al., 2015). MS is less prevalent in non-white


European populations, (Zhuange et al., 2015) but Wallin and colleagues (Wallin et al.,
2012) found that the epidemiological landscape may be changing— its prevalence
among Black military members was nearly 1.5 times greater than Whites from the same
group (12.1 incidents per 100,000 individuals per year vs. 9.3, respectively. The rate for
Hispanics was 8.3). Examining the genetic determinants of MS susceptibility in groups
with different ancestry could provide important clues to heterogeneity in MS risk
worldwide (Westerlind et al., 2015).
In conclusion, this study shows significant pleiotropy between genomic modulators of
MS risk and regional brain volumes, in the hippocampus and thalamus. We found two
missense candidate variants related to both brain structure (thalamic and hippocampal
volumes) and MS risk. Future studies are needed in order to evaluate the importance of
these loci. Specifically, a useful study design might be to test for associations between
the MS-risk loci identified in our analysis and both MS brain lesion burden and ensuing
phenotypic changes in patients. We identified 22 other putative MS susceptibility
variants not identified in previous studies, 16 were at least moderately associated with
those concurrently discovered independently by the IMSGC in a genetic association
study. Here we show the utility of CIA, conditional FDR and contribute new knowledge
of the genetic architecture of MS.
21
2.5 Chapter 2 References
1000 Genomes Project Consortium. A global reference for human genetic variation.
Nature. 2015 Oct 1;526(7571):68.
Anaya JM, Kim-Howard X, Prahalad S, Cherñavsky A, Cañas C, Rojas-Villarraga A,
Bohnsack J, Jonsson R, Bolstad AI, Brun JG, Cobb B. Evaluation of genetic association
between an ITGAM non-synonymous SNP (rs1143679) and multiple autoimmune
diseases. Autoimmunity reviews. 2012 Feb 29;11(4):276-80.
Andreassen OA, Djurovic S, Thompson WK, Schork AJ, Kendler KS, O’Donovan MC,
Rujescu D, Werge T, van de Bunt M, Morris AP , McCarthy MI. Improved detection of
common variants associated with schizophrenia by leveraging pleiotropy with
cardiovascular-disease risk factors. The American Journal of Human Genetics. 2013
Feb 7;92(2):197-209.
Andreassen OA, Harbo HF , Wang Y, Thompson WK, Schork AJ, Mattingsdal M, Zuber V,
Bettella F , Ripke S, Kelsoe JR, Kendler KS. Genetic pleiotropy between multiple
sclerosis and schizophrenia but not bipolar disorder: differential involvement of
immune-related gene loci. Molecular psychiatry. 2015 Feb 1;20(2):207-14.
Ascherio A. Environmental factors in multiple sclerosis. Expert review of
neurotherapeutics. 2013 Dec 1;13(sup2):3-9.
Baranzini SE. Revealing the genetic basis of multiple sclerosis: are we there yet?
Current opinion in genetics & development. 2011 Jun 30;21(3):317-24.
Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful
approach to multiple testing. Journal of the Royal Statistical Society. Series B
(Methodological). 1995 Jan 1:289-300.
Bermel RA, Bakshi R. The measurement and clinical relevance of brain atrophy in
multiple sclerosis. The Lancet Neurology. 2006 Feb 28;5(2):158-70.
Bulik-Sullivan BK, Loh PR, Finucane HK, Ripke S, Yang J, Patterson N, Daly MJ, Price
AL, Neale BM, Schizophrenia Working Group of the Psychiatric Genomics Consortium.
LD Score regression distinguishes confounding from polygenicity in genome-wide
association studies. Nature genetics. 2015 Mar 1;47(3):291-5.
Endele S, Zabel B, Winterpacht A. Assignment of the human diacylglycerol kinase 4
(DAGK4) gene to chromosome 4p16. 3. Genomics. 1996 Apr 1;33(1):145-6.
Eyler LT, Prom‐Wormley E, Fennema‐Notestine C, Panizzon MS, Neale MC, Jernigan
TL, Fischl B, Franz CE, Lyons MJ, Stevens A, Pacheco J. Genetic patterns of correlation
22
among subcortical volumes in humans: results from a magnetic resonance imaging twin
study. Human brain mapping. 2011 Apr 1;32(4):641-53.
Geurts JJ, Barkhof F . Grey matter pathology in multiple sclerosis. The Lancet Neurology.
2008 Sep 30;7(9):841-51.
Gourraud PA, Sdika M, Khankhanian P , Henry RG, Beheshtian A, Matthews PM, Hauser
SL, Oksenberg JR, Pelletier D, Baranzini SE. A genome-wide association study of brain
lesion distribution in multiple sclerosis. Brain. 2013 Feb 13;136(4):1012-24.
Hassan-Smith G, Douglas MR. Epidemiology and diagnosis of multiple sclerosis. British
journal of hospital medicine (London, England: 2005). 2011 Oct;72(10):M146-51.
Hibar DP , Jahanshad N, Medland SE, Thompson PM. Continuous inflation analysis: a
threshold-free method to estimate genetic overlap and boost power in imaging
genetics. MICCAI Imaging Genetics Workshop, 2015, IN PRESS.
Hibar DP , Stein JL, Renteria ME, Arias-Vasquez A, Desrivières S, Jahanshad N, Toro R,
Wittfeld K, Abramovic L, Andersson M, Aribisala BS. Common genetic variants
influence human subcortical brain structures. Nature. 2015 Apr 9;520(7546):224-9.
International Multiple Sclerosis Genetics Consortium, Wellcome Trust Case Control
Consortium 2. Genetic risk and a primary role for cell-mediated immune mechanisms in
multiple sclerosis. Nature. 2011 Aug 11;476(7359):214-9.
International Multiple Sclerosis Genetics Consortium. Analysis of immune-related loci
identifies 48 new susceptibility variants for multiple sclerosis. Nature genetics. 2013 Nov
1;45(11):1353-60.
International Multiple Sclerosis Genetics Consortium., Hafler DA, Compston A, Sawcer
S, Lander ES, Daly MJ, De Jager PL, de Bakker PI, Gabriel SB, Mirel DB, Ivinson AJ,
Pericak-Vance MA, Gregory SG, Rioux JD, McCauley JL, Haines JL, Barcellos LF , Cree
B, Oksenberg JR, Hauser SL. Risk alleles for multiple sclerosis identified by a
genomewide study. N . Risk alleles for multiple sclerosis identified by a genomewide
study. N engl J med. 2007 Aug 30;357:851-62
Isobe N, Madireddy L, Khankhanian P , Matsushita T, Caillier SJ, Moré JM, Gourraud PA,
McCauley JL, Beecham AH, Piccio L, Herbert J. An ImmunoChip study of multiple
sclerosis risk in African Americans. Brain. 2015 Mar 27:awv078.
Lauc G, Huffman JE, Pučić M, Zgaga L, Adamczyk B, Mužinić A, Novokmet M, Polašek
O, Gornik O, Krištić J, Keser T. Loci associated with N-glycosylation of human
immunoglobulin G show pleiotropy with autoimmune diseases and haematological
cancers. PLoS Genet. 2013 Jan 31;9(1):e1003225.
23
Lehmann PV, Rottlaender A, Kuerten S. The autoimmune pathogenesis of multiple
sclerosis. Die Pharmazie-An International Journal of Pharmaceutical Sciences. 2015 Jan
12;70(1):5-11.
Liu Y, Duan Y, Huang J, Ren Z, Ye J, Dong H, Shi FD, Barkhof F , Vrenken H, Wattjes MP ,
Wang J. Multimodal Quantitative MR Imaging of the Thalamus in Multiple Sclerosis and
Neuromyelitis Optica. Radiology. 2015 Jun 4;277(3):784-92
Miller DH, Barkhof F , Frank JA, Parker GJ, Thompson AJ. Measurement of atrophy in
multiple sclerosis: pathological basis, methodological aspects and clinical relevance.
Brain. 2002 Aug 1;125(8):1676-95.
Moutsianas L, Jostins L, Beecham AH, Dilthey AT, Xifara DK, Ban M, Shah TS,
Patsopoulos NA, Alfredsson L, Anderson CA, Attfield KE, Baranzini SE, Barrett J, Binder
TM, Booth D, Buck D, Celius EG, Cotsapas C, D'Alfonso S, Dendrou CA, Donnelly P ,
Dubois B, Fontaine B, Lar Fugger L, Goris A, Gourraud PA, Graetz C, Hemmer B, Hillert
J; International IBD Genetics Consortium (IIBDGC)., Kockum I, Leslie S, Lill CM,
Martinelli-Boneschi F , Oksenberg JR, Olsson T, Oturai A, Saarela J, Søndergaard HB,
Spurkland A, Taylor B, Winkelmann J, Zipp F , Haines JL, Pericak-Vance MA, Spencer
CC, Stewart G, Hafler DA, Ivinson AJ, Harbo HF , Hauser SL, De Jager PL, Compston A,
McCauley JL, Sawcer S, McVean G; International Multiple Sclerosis Genetics
Consortium. Class II HLA interactions modulate genetic risk for multiple sclerosis. Nat
Genet. 2015 Oct;47(10):1107-13.
Nagase T, Seki N, Tanaka A, Ishikawa KI, Nomura N. Prediction of the coding
sequences of unidentified human genes. IV. The coding sequences of 40 new genes
(KIAA0121-KIAA0160) deduced by analysis of cDNA clones from human cell line KG-1.
DNA Research. 1995 Jan 1;2(4):167-74.
Nyholt DR. SECA: SNP effect concordance analysis using genome-wide association
summary results. Bioinformatics. 2014 Apr 1:btu171.
Patsopoulos NA, Barcellos LF , Hintzen RQ, Schaefer C, Van Duijn CM, Noble JA, Raj T,
Gourraud PA, Stranger BE, Oksenberg J, Olsson T. Fine-mapping the genetic
association of the major histocompatibility complex in multiple sclerosis: HLA and
non-HLA effects. PLoS Genet. 2013 Nov 21;9(11):e1003926.
Patsopoulos NA, Baranzini SE, Santaniello A, Shoostari P , Cotsapas C, Wong G,
Beecham AH, James T, Replogle J, Vlachos I, McCabe C. The Multiple Sclerosis
Genomic Map: Role of peripheral immune cells and resident microglia in susceptibility.
bioRxiv. 2017 Jan 1:143933.
24
Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, Maller J, Sklar P ,
De Bakker PI, Daly MJ, Sham PC. PLINK: a tool set for whole-genome association and
population-based linkage analyses. The American Journal of Human Genetics. 2007
Sep 30;81(3):559-75.
Schmiesing JA, Gregson HC, Zhou S, Yokomori K. A human condensin complex
containing hCAP-C–hCAP-E and CNAP1, a homolog of Xenopus XCAP-D2, colocalizes
with phosphorylated histone H3 during the early stage of mitotic chromosome
condensation. Molecular and cellular biology. 2000 Sep 15;20(18):6996-7006.
Sicotte NL, Kern KC, Giesser BS, Arshanapalli A, Schultz A, Montag M, Wang H,
Bookheimer SY. Regional hippocampal atrophy in multiple sclerosis. Brain. 2008 Apr
1;131(4):1134-41.
Simon DI, Chen Z, Seifert P , Edelman ER, Ballantyne CM, Rogers C. Decreased
neointimal formation in Mac-1–/–mice reveals a role for inflammation in vascular repair
after angioplasty. The Journal of clinical investigation. 2000 Feb 1;105(3):293-300.
Sivakumaran S, Agakov F , Theodoratou E, Prendergast JG, Zgaga L, Manolio T, Rudan
I, McKeigue P , Wilson JF , Campbell H. Abundant pleiotropy in human complex diseases
and traits. The American Journal of Human Genetics. 2011 Nov 11;89(5):607-18.
Wallin MT, Culpepper WJ, Coffman P , Pulaski S, Maloni H, Mahan CM, Haselkorn JK,
Kurtzke JF , Veterans Affairs Multiple Sclerosis Centres of Excellence Epidemiology
Group. The Gulf War era multiple sclerosis cohort: age and incidence rates by race, sex
and service. Brain. 2012 Jun 1;135(6):1778-85.
Wasco W, Brook JD, Tanzi RE. The amyloid precursor-like protein (APLP) gene maps to
the long arm of human chromosome 19. Genomics. 1993 Jan 31;15(1):237-9.
Westerlind H, Imrell K, Ramanujam R, Myhr KM, Celius EG, Harbo HF , Oturai AB,
Hamsten A, Alfredsson L, Olsson T, Kockum I. Identity-by-descent mapping in a
Scandinavian multiple sclerosis cohort. European Journal of Human Genetics. 2015
May 1;23(5):688-92.
Zhuang JC, Huang ZY, Zhao GX, Yu H, Li ZX, Wu ZY. Variants of CYP27B1 are
associated with both multiple sclerosis and neuromyelitis optica patients in Han
Chinese population. Gene. 2015 Feb 25;557(2):236-9.
Zubenko GS, Hughes HB. Predicted gene sequence C10orf112 is transcribed, exhibits
tissue-specific expression, and may correspond to AD7. Genomics. 2009 Apr
30;93(4):376-82.
25
CHAPTER THREE
Genetic pleiotropy between determinants of
coronary artery disease risk and regional brain volumes
3.1 Introduction
See Chapter 1 for a more detailed introduction to coronary artery disease epidemiology
and genetics.
The degree to which shared genetic factors influence CAD and brain morphology will
give important insight to the pathogenesis of CAD. Here, using SNP Effect
Concordance Analysis (SECA) we combined summary statistics from the CARDIoGRAM
GWAS with those from Hibar et al. 2015 GWAS of brain volumes, we tested for (1)
global genetic pleiotropy (2) genetic concordance between genetic variants associated
with CAD and brain volume.
3.2 Methods
3.2.1 Description of original association studies
We analyzed summary statistics from the publicly available large-scale publicly available
GWASs of coronary artery disease (CAD), by CARDIoGRAM (34,997 cases and 49,512
controls) (Nikpay et al., 2015), and of brain volumes, by the ENIGMA Consortium
(13,171 subjects across 50 cohorts worldwide) (Hibar et al., 2015). The ENIGMA
meta-analyses consist of separate GWASs of seven subcortical brain volumes (nucleus
accumbens, amygdala, caudate nucleus, hippocampus, globus pallidus, putamen,
thalamus) and total intracranial volume (ICV). No participants in the ENIGMA GWASs
were diagnosed with CAD at the time of scanning (Hibar et al., 2015). All subjects from
both studies were of European ancestry, which was verified by multidimensional scaling
(MDS) or principal component analysis (PCA). GWAS summary statistics were
gnome-controlled to adjust for spurious inflation.
3.2.2 Post-processing of genetic data
After quality control and filtering of the CAD data, 2,355,027 SNPs remained (see
Schunkert et al., 2011, for site-level imputation and quality control details). For the eight
brain volume GWASs, 8,398,366 SNPs were available after quality control and filtering
(see Hibar et al., 2015, Methods for imputation and quality control details). 1,933,008
SNPs passed quality control and filtering restrictions across all datasets.
Using PLINK (Purcell et al., 2007), we performed a clumping procedure to identify an
independent SNP from each linkage disequilibrium (LD) block across the genome to
26
create sets of independent SNPs for the datasets on which we are conditioning the
CAD GWAS. Thus, for each brain volume GWAS, we applied this clumping process
using a 500 Kb window with SNPs in LD (r
2
> 0.25) in the European reference samples
from the 1000 Genomes Project (Phase 1, version 3). To produce unbiased estimates of
concordance and pleiotropy, we included the entire genome in the clumping procedure
by setting the inclusion threshold for both indexed and clumped SNPs to P ≤ 1. For
each LD block, the SNP with the lowest p-value was selected as an index SNP
representing that block, and all other SNPs in that block were excluded. Clumping
yielded eight sets of independent SNPs (one for each brain volume GWAS) that
represent the total variation explained by the entire genome, conditioned on the
significance in each brain volume GWAS. For each set of brain volume SNPs, we
extracted the same set of SNPs and corresponding summary statistics from the CAD
GWAS for subsequent analyses.
3.2.3 Tests of pleiotropy and concordance
We used SNP Effect Concordance Analysis (SECA) (Nyholt, 2014). to assess the overlap
between common genetic variants associated with both CAD and a given brain volume
measure. To measure pleiotropy, we ordered SNPs based on their p-value for
association for each brain volume and iterated through 12 p-value thresholds (0.01,
0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1) and counted how many SNPs
overlapped between the two traits at each threshold. We used a binomial test to see if
the number of overlapping SNPs was greater than expected at random, under the null
hypothesis of no pleiotropy. We compared all levels of the brain volume GWAS and all
levels of the CAD GWAS (144 comparisons); then, we determined the number of
comparisons with evidence for pleiotropy at nominal significance (P ≤ 0.05). To estimate
global pleiotropy, we generated 10,000 permuted datasets for each brain volume-CAD
pair and determined if the number of significant thresholds was significantly greater
than expected by chance.
We also used SECA to estimate concordance, a measure of agreement of directions in
SNP effects between two traits. We used a two-sided Fisher’s exact test to check for a
nominally significant (P ≤ 0.05) positive or negative trend in effect sizes of overlapping
SNPs. Then, we estimated global concordance by generating 10,000 permuted
datasets and performing the Fisher’s exact test, assessing whether the number of
significant overlapping thresholds was significantly greater than expected (see Nyholt et
al., 2014 for details of the SECA analysis).
We tested for pleiotropy and concordance between CAD and each of the seven
subcortical structures and ICV . Applying the Bonferroni correction for the number of
tests, we set the study-wide significant threshold to P* = 0.05 / (2 * 8) = 3.13 x 10
-3
.
27

3.3 Results

3.3.1 Pleiotropy between globus pallidus, nucleus accumbens, ICV, and CAD-associated
variants

Using SECA, we found evidence of global pleiotropy between CAD risk variants and
variants associated with globus pallidus (P = 0.000999), nucleus accumbens (P =
0.000999), and ICV (P = 0.002) (Figure 3.1). The other five structures did not show
study-wide significant pleiotropy with CAD: thalamus (P = 0.004), amygdala (P =
0.00799), hippocampus (P = 0.023), caudate (P = 0.0689), and putamen (P = 0.311).
Plotted outputs from the global pleiotropy test are shown in Figure 3.1

Figure 3.1 SECA binomial test heatmaps showing evidence of pleiotropy
between CAD risk conditioned on genetic determinants of globus pallidus,
nucleus accumbens, and ICV, respectively from left to right. In each plot,
the x-axis represents SNPs associated with CAD risk ranked by p-value,
the y-axis represents each results from each respective brain volume
GWAS, ranked by p-value.

3.3.2 Concordance between ICV and CAD-associated variants

We found concordance (same SNP , same effect direction)
between CAD risk variants and variants associated with
a decrease in ICV (P = 0.000999). For the remaining seven
structures, we did not observe study-wide significant
evidence of concordance in either direction: amygdala (P =
0.012), caudate (P = 0.023), globus pallidus (P = 0.028),
thalamus (P = 0.036), hippocampus (P = 0.244), putamen
(P > 0.99), and nucleus accumbens (P > 0.99).
Figure 3.2 SECA binomial test plot showing concordance between CAD risk and
reduced ICV.

28

3.4 Discussion

To our knowledge, this is the first study to directly test and demonstrate a concrete
genetic link between CAD risk and quantitative MRI derived brain morphology. Our
findings complement a substantial body of literature reporting overlap between CAD
risk factors and neurocognitive dysfunction.

In comparing results from the two largest genetic studies of CAD risk and brain
morphology, we found strong evidence of genetic overlap between the two traits.
Specifically genetic modulators of globus pallidus and nucleus accumbens volume, in
addition to total intracranial volume, were found to be significantly pleiotropic with CAD
risk using the SECA method.

We additionally tested for genetic concordance between brain volume and CAD risk
using SECA, testing for agreement in the direction of effect between the two compared
traits. We found discordance between CAD risk variants and ICV modulating variants,
indicating that genetic variants that influence a decrease in ICV also increase CAD risk.

29

3.5 Chapter 3 References

Bulik-Sullivan, B., et al. An atlas of genetic correlations across human diseases and
traits. Nat Genet 47, 1236-1241 (2015).

Bulik-Sullivan, B.K., et al. LD Score regression distinguishes confounding from
polygenicity in genome-wide association studies. Nat Genet 47, 291-295 (2015).

Eggermont LHP , de Boer K, Muller M, et al Cardiac disease and cognitive impairment: a
systematic review Heart 2012;98:1334-1340.

Gąsecki D, Kwarciany M, Nyka W, Narkiewicz K. Hypertension, brain damage and
cognitive decline. Current hypertension reports. 2013 Dec 1;15(6):547-58.

Go AS, Mozaffarian D, Roger VL, et al. Heart disease and stroke statistics—2013 update: a
report from the American Heart Association. Circulation. 2013;127:e6–e245.

Hibar, D.P ., et al. Common genetic variants influence human subcortical brain structures. Nature
520, 224-229 (2015).

McEvoy LK, Fennema-Notestine C, Eyler LT, Franz CE, Hagler DJ, Lyons MJ, Panizzon
MS, Rinker DA, Dale AM, Kremen WS. Hypertension-Related Alterations in White Matter
Microstructure Detectable in Middle AgeNovelty and Significance. Hypertension. 2015
Aug 1;66(2):317-23.

McPherson R, Tybjærg-Hansen A. Genetics of Coronary Artery Disease. Circulation
research. 2016 Feb;118(4):564-78.

Nikpay, M., Goel, A., Won, H.-H., Hall, L. M., Willenborg, C., Kanoni, S., … Farrall, M.
(2015). A comprehensive 1000 Genomes-based genome-wide association
meta-analysis of coronary artery disease. Nature Genetics, 47(10), 1121–1130.

Nyholt, D.R. SECA: SNP effect concordance analysis using genome-wide association
summary results. Bioinformatics 30, 2086-2088 (2014).

Purcell, S., et al. PLINK: a tool set for whole-genome association and population-based
linkage analyses. Am J Hum Genet 81, 559-575 (2007).

Roberts, R. (2014). Genetics of Coronary Artery Disease: An Update. Methodist
DeBakey Cardiovascular Journal, 10(1), 7–12.

Schunkert, H., et al. Large-scale association analysis identifies 13 new susceptibility loci
for coronary artery disease. Nat Genet 43, 333-338 (2011).

30

The European health report 2012: charting the way to well-being. Copenhagen,
Denmark: WHO Regional Office for Europe; 2012.

Vogels R, L, C, Oosterman J, M, van Harten B, Gouw A, A, Schroeder-Tanka J, M,
Scheltens P , van der Flier W, M, Weinstein H, C, Neuroimaging and Correlates of
Cognitive Function among Patients with Heart Failure. Dement Geriatr Cogn Disord
2007;24:418-423

Vogels RL, Scheltens P , Schroeder-Tanka JM, Weinstein HC: Cognitive impairment in
heart failure: a systematic review of the literature. Eur J Heart Fail 2007;9:440–449.


31

CHAPTER FOUR
Multiple sclerosis risk variants associated with DTI white-matter integrity

4.1 Introduction

As described in Chapter 1, MS is a chronic neurodegenerative disease with an unknown
pathogenesis. It is a complex polygenic disorder (Bush et al., 2010), where genetic
predisposition plays an important role in disease risk. It is highly heterogeneous, with
vast phenotypic variation and progression between patients (Hassan-Smith and
Douglas, 2011). The source of this mysterious heterogeneity has been hypothesized to
be the underlying genetic contribution to disease risk (Jersild, Svejgaard, and Fog,
1972; Longbrake and Hafler, 2016).

Recent large-scale genome-wide association (GWA) studies and follow-up efforts have
identified over 200 common genetic variants significantly associated with disease risk,
explaining ~39% of the disease heritability (Patsopoulos, et al., 2017, Baranzini et al.,
2017). How these risk variants influence disease susceptibility is not well-understood.
Despite a clear need, MS studies relating genetic risk to clinical phenotype are scarce,
especially MRI (Isobe et al., 2016), and significant findings in this realm have been
described in the field as “the MS holy grail.” (Longbrake and Hafler, 2016).

RPS6KB1 is a confirmed MS risk locus (Sawcer et al., 2011, Lill et al., 2013). It encodes
the protein p70s6k, which is a part of the mTOR signaling pathway. It known to directly
affect MS as a regulator of inflammatory response in microglia and as an actor in
remyelination after CNS injury (Abdollah Zadeh et al., 2017). rs180515, in particular,
modulates vitamin D levels. Vitamin D has been one of the most widely documented
environmental risk factors in MS (Ascherio et al., 2014).

Multiple studies have implicated CD6 as an MS protective gene (Swaminathan
et al., 2013; D'Cunha et al., 2016, Li et al., 2017), and it is a known factor in other
autoimmune diseases (Da Gloria et al., 2014). CD6 KO mice are protected from
developing experimental autoimmune encephalomyelitis (EAE), the mice model of MS.
CD6 is a T cell surface protein and helps to enable the transport of leukocytes through
the blood brain barrier (BBB) (Li et al., 2017, Wagner et al., 2014). Furthermore, Wagner
et al., found differences in mRNA expression between MS patients versus controls. The
gene has been shown to interact with another MS risk gene ALCAM, in one of the first
studies to test gene-gene interactions in MS.

The hallmark symptoms of MS are caused by demyelination and subsequent damage to
neural tissue in the CNS. This damage is routinely assessed in the diagnosis of MS
using clinical MRI—standard practice for over 20 years (Polman et al., 2011)—and in
clinical trials as an outcome measure to assess treatment efficacy (Cutter and Kappos,
2014). Lesions detected with MRI generally represent tissue damage that has been
32

developing for some time. An MS diagnosis often occurs long after a prodromal period,
which, in some cases can be significantly protracted. Additionally, tissue damage
occurs outside of macroscopic lesions in normal appearing gray and white matter
(NAGM, NAWM), known as occult injury. Using conventional MRI in MS research, we
are limited to imaging the macrostructural vestiges of a microstructural disease.

Diffusion tensor MRI (DTI) has been integral in understand MS pathology in vivo as it
provides surrogate markers of WM microstructure and can be used to indirectly assess
demyelination. Differences in DTI-derived measurements have been shown to predict
new lesion formation, although the effect is subtle (Werring et al., 2000; Inglese and
Bester, 2010).

While predominantly in the WM, demyelination in MS occurs throughout the CNS in
both WM and GM. It is suspected that different mechanisms are responsible for the
targeting of different tissues. Demyelinating factors likely arrive through inflammatory
cells or directly from the ventricular system are suspected to be a source of lesions in
subcortical GM structures. (Gilmore et al., 2009).

It is known that MS variants affect gene expression in brain tissue, definitely in the
cortex, however it is reasonable to assume this is true elsewhere in the CNS, such as
cerebral white-matter (Patsopoulos, 2017).

Here we explore the effect of two MS risk variants —rs180515 on RPS6KB1 and
rs650258 on CD6 —on brain imaging measures. We then overlaid our statistical result
maps with whole-brain gene expression data from the Allen Brain Atlas.

4.2 Methods

4.2.1 Subjects, genotyping and image acquisition information.

A detailed description of the participant cohort is described in Braskie et al., 2011.
Briefly, 398 (mean age, 23.6 +/- 2.2 years; age range, 20–29 years) healthy young
Australian twins as part of the Queensland Twin Imaging Study. This group of 398
passed exclusionary criteria of ancestry, scan quality, and good health. The
Human610-Quad BeadChip (Illumina) was used to obtain genomic DNA information, per
manufacturer’s protocols (Infinium HD Assay; Super Protocol Guide; Rev. A, May 2008).
Genomic quality control was employed and the sample here was within normal
Hardy-Weinberg equilibrium. The study met standards set by the National Statement on
Ethical Conduct in Human Research (2007) issued by the National Health and Medical
Research Council of Australia and was approved by the Queensland Institute of Medical
Research Human Research Ethics Committee.
4.2.2 Diffusion tensor imaging (DTI)
33

Detailed imaging parameters are described in Braskie et al., 2011. T1-weighted brain
images were acquired with gradient echo sequence on a 4 Tesla MRI scanner (Bruker
Medspec; TI/TR/TE = 700/1500/3.35 ms; flip angle =8°; slice thickness = 0.9 mm, 256 ×
256 × 256). Single-shot echo planar imaging was used to acquire diffusion-weighted
images, with twice-refocused spin echo to enable eddy-current induced distortion
correction (TR/TE 6090/91.7 ms, 23 cm FOV, 128×128). 55 axial slices comprised each
3d volume (2.0 mm/0 mm gap; 1.79 mm ×1.79 mm in-plane resolution). 105 images per
subject: 11 with no diffusion sensitization (i.e., T2-weighted b0 images) were acquired
as well as 94 diffusion-weighted (DW) images (b = 1149 s/mm2) with gradient directions
evenly distributed on the hemisphere.

4.2.3 DTI processing

Non-brain regions from all images were automatically removed using FSL BET
(http://fsl.fmrib.ox.ac.uk/fsl/) and manually edited extraction. T1 images were lineraly
aligned to a common space. DW images were eddy current distortion corrected using
FSL “eddy_correct” method (http://fsl.fmrib.ox.ac.uk/fsl/). DW b0 images were
averaged and linearly aligned and resampled to corresponding T1 images. Standard
control for EPI-induced artifacts were employed. FA values were calculated at each
voxel as described in Braskie et al. 2011.

4.2.4 Statistical analysis

Genetic association analysis was conducted using a mixed-model regression at each
white matter voxel in FA maps (Zhou et al., 2012). Voxelwise associations controlled for
age, sex and accounted for kinship. The widely used False Discovery Rate method
(Benjamini and Hochberg 1995) was employed to control for multiple comparisons.

4.2.5 Preliminary Replication cohort

To assess our results in another cohort, we used the publicly available Alzheimer’s
Disease Neuroimaging Initiative phase 1. ADNI-1 MRI dataset which contains 802 (184
AD, 391 MCI, 229 controls) T1 brain images. This dataset is described in detail in Wang
et al., 2011. At the time we ran this experiment, DTI information was not available in the
ADNI-1 dataset, so we used tensor based morphometry (TBM) to measure structural
brain differences across subjects. We employed linear regression here for voxelwise
tests of SNP effects. Our statistical model also controlled for age and sex. Searchlight
FDR (Langers et al, 2007) was used to correct for multiple comparisons.

4.2.6 The Allen Human Brain Atlas gene expression analysis
34

The Allen Human Brain Atlas (http://human.brain-map.org) provides publically available
RNA microarray data from postmortem brains of 6 donors with no indications of
neuropathology or psychiatric diagnoses. The atlas provides a searchable index of
spatially normalized expression profiles of 29,191 genes represented by 58,692 probes
(ALLEN Human Brain Atlas Normalization, Microarray Data, 2013).

We used the AllenBrainPy software package (Roshchupkin, 2016,
http://github.com/roshchupkin/AllenBrainPy), to extract gene expression profiles of five
donors of the Allen Human Brain Atlas. Gene expression maps are represented in MNI
coordinate space and z-scores were generated to identify clusters of greater than
background gene expression. These maps were then registered and overlaid with
white-matter integrity association results maps with clusters of p<0.01 significance. We
then generated t-test statistics to assess whether gene expression for a given variant
(RPS6KB1, CD6) was found significantly more in the FA association clusters. Bonferroni
correction was employed to address multiple comparisons. Further information as to
how this data was obtained can be found in Roschchupkin et al, 2016.

4.3 Results

Here we assessed white matter integrity using fractional anisotropy from DTI scans of
398 healthy young adults (mean age, 23.6 ± 2.2 years), who were also genotyped for
each SNP in RPS6KB1 and CD6 genes. We also tested a replication sample of 842
healthy adults.

In statistical maps generated from the 398 young adults, each G allele for rs650258 of
CD6 was significantly associated with higher FA in the cingulum, bilateral SLF , forceps
major, left ILF and other unspecified regions of white matter, at the FDR critical P value
of 0.004. The G allele of rs650258 was also associated with structural brain differences
as measured by TBM in the ADNI-1 dataset.

Similarly, each G allele for rs180515 SNP of RPS6KB was also associated with higher
FA. Again, several clusters of voxels survived the FDR critical P of 0.004. These regions
were in the superior region of the corona radiata, superior thalamic radiation, ILF and
other white matter regions. Again, this finding was evaluated in TBM analysis with
ADNI-1 subjects.

Results are displayed in Figures 4.1 - 4.5.
35

Figure 4.1. Statistical maps of rs180515 association with FA in 398 healthy young adults. Results are overlaid onto
average FA maps. All colored regions pass FDR critical P value .0043, minP = 6.571e-06. Colors indicate regression β
values.

Figure 4.2.
Replication of rs180515 in ADNI dataset. Statistical maps of rs180515 association with TBM in 842 healthy adults..
Results are overlaid onto average structural T1 images. All colored regions pass FDR critical P value .05, minP = .03613
Colors indicate regression β values.

36

For all 5 donors, we found that the MS risk gene, RPS6KB1, was significantly expressed
in voxel clusters that were previously shown to display significantly higher FA in carriers
of an allele from that gene (min p=7.11x10-11).

Figure 4.3. Overlay of Allen Brain Atlas RPS6KB1 gene expression maps with results from white-matter association
study. Colored regions are significant at FDR critical value. Color bars below images show gene expression z-score by
brain region.

Figure 4.4. Statistical maps of rs650250 association with FA in 398 healthy young adults. Results are overlaid onto
average FA maps. All colored regions pass FDR critical P value .0036 minP = 2.011e-07. Colors indicate regression β
values.

37

Figure 4.5. Replication of rs650258 in ADNI-1I. Statistical maps of rs650258 association with TBM in 842 healthy adults.
Results are overlaid onto average structural T1 images. All colored regions pass FDR critical P value .05, minP = .0318
Colors indicate regression β values.

In 3/5 donors, CD4, was significantly expressed in associated voxels (min p=1.7x10-4).

Figure 4.6. Overlay of Allen Brain Atlas RPS6KB1 gene expression maps with results from white-matter association
study. Colored regions are significant at FDR critical value. Color bars below images show gene expression z-score by
brain region.

38

4.4 Discussion

We found higher FA in the white matter of participants who lacked the protective allele
rs180515 on the CD6 gene or carried the rs650258 variant on RP6KB1. While
single-SNP association studies have well-documented limitations, it is of note that
these findings passed stringent multiple comparisons correction and we replicated
these findings in an independent dataset from the ADNI cohort; and additionally, found
significant overlap with gene expression maps provided by the Allen Brain Atlas.

Both of these variants were originally identified in the 2011 IMSGC MS GWAS and have
subsequently been replicated and confirmed as true MS risk variants (Sawcer et al.,
2011, Lill et al., 2013, Abdollah Zadeh et al., 2017; Swaminathan
et al., 2013; D'Cunha et al., 2016, Li et al., 2017).

In classic imaging genetics studies—combining neuroimaging derived measures with
genetic data—results from single-SNP association should be taken with extreme
caution. Linking a gene directly to a phenotype, especially in a polygenic disease, is
highly complex and massive power is required to find true associations (Gottesman and
Gould, 2003; Bearden and Freimer, 2006; Cannon and Keller, 2006).

It is of note that the effect sizes (β from regression) of our results are very small (max <
.1) for all SNP-MRI associations. This is likely due to the fact that these variants—like
most detected by GWAS—explain a very small amount of the variance in disease
association, even less so for a particular phenotype. These small effects are the
“rate-limiting factor” in gene discovery (Meyer-Lindenberg and Weinberger, 2006).

Most published single SNP candidate gene association studies fail to replicate.
(Ioannidis, Tarone, & McLaughlin, 2011; Ioannidis, 2005). Meta-analyses on these
studies, usually agree that significant findings are mostly overestimated. Most SNPs in
complex diseases confer odds ratios of 1.1-1.5 (10-50% relative increase in probability
of developing a disease), as is the case in our study, which explains at most 8% of
disease risk at a population level (Ioannidis, 2003), increasing the difficulty of
establishing findings in a small healthy sample. Further, the combinatorial architecture
of genes makes it difficult to establish a functional risk with a single genetic variant.
Even if a given SNP theoretically explains a large part of the variance of a disease, it
may still rely on epistasis to function properly (Hill et al., 2008).

It should be noted that our replication sample contained 184 AD patients and 391 MCI
subjects. We chose to include the full sample including AD patients to maximize power.
While this increased the variance of morphometry in our sample, it is not certain that
this would influence our results. While both diseases are defined inflammation and
demyelination, research suggests MS related pathology has no influence on Alzheimer’s
(Dal Bianco, et al., 2008).
39

Future work is required to examine complex putative gene-gene interactions. An
experimental attempt to improve upon the single-SNP technique by summarizing
genetic predisposition with polygenic risk scoring is described in the next chapter.
40

4.5 Chapter 4 References

Abdollah Zadeh R, Jalilian N, Sahraian MA, Kasraian Z, Noori-Daloii MR.
Polymorphisms of RPS6KB1 and CD86 associates with susceptibility to multiple
sclerosis in Iranian population. Neurological research. 2017 Mar 4;39(3):217-22.

Ascherio A, Munger KL, White R, Köchert K, Simon KC, Polman CH, Freedman MS,
Hartung HP , Miller DH, Montalbán X, Edan G. Vitamin D as an early predictor of multiple
sclerosis activity and progression. JAMA neurology. 2014 Mar 1;71(3):306-14.
Baranzini SE, Oksenberg JR. The Genetics of Multiple Sclerosis: From 0 to 200 in 50
Years. Trends in Genetics. 2017 Oct 5.

Bearden CE, Freimer NB. Endophenotypes for psychiatric disorders: ready for
primetime? Trends Genet. 2006 Jun;22(6):30613. Epub 2006 May 11. PubMed PMID:
16697071.

Benjamini Y , Hochberg Y. Controlling the false discovery rate: a practical and powerful
approach to multiple testing. Journal of the Royal Statistical Society. Series B
(Methodological). 1995 Jan 1:289-300.

Braskie MN, Jahanshad N, Stein JL, Barysheva M, McMahon KL, de Zubicaray GI,
Martin NG, Wright MJ, Ringman JM, Toga AW, Thompson PM. Common Alzheimer's
disease risk variant within the CLU gene affects white matter microstructure in young
adults. Journal of Neuroscience. 2011 May 4;31(18):6764-70.

Bush WS, Sawcer SJ, de Jager PL, Oksenberg JR, McCauley JL, Pericak-Vance MA,
Haines JL. International Multiple Sclerosis Genetics Consortium (IMSGC).(2010).
Evidence for polygenic susceptibility to multiple sclerosis—the shape of things to come.
Am. J. Hum. Genet.;86:621-5.

Cannon TD, Keller MC. Endophenotypes in the genetic analyses of mental
disorders. Annu Rev Clin Psychol. 2006;2:26790. Review. PubMed PMID: 17716071.

Cutter GA, Kappos LU. Clinical trials in multiple sclerosis. Handb Clin Neurol. 2014 Feb
5;122:445-53.

D’Cunha MA, Pandit L, Malli C. CD6 gene polymorphism rs17824933 is associated with
multiple sclerosis in Indian population. Annals of Indian Academy of Neurology. 2016
Oct;19(4):491.

da Glória VG, de Araújo MM, Santos AM, Leal R, de Almeida SF , Carmo AM, Moreira A.
T cell activation regulates CD6 alternative splicing by transcription dynamics and
SRSF1. The Journal of Immunology. 2014 Jul 1;193(1):391-9.
41

Dal Bianco A, Bradl M, Frischer J, Kutzelnigg A, Jellinger K, Lassmann H. Multiple
sclerosis and Alzheimer's disease. Annals of neurology. 2008 Feb 1;63(2):174-83.

Gilmore CP , Geurts JJ, Evangelou N, Bot JC, Van Schijndel RA, Pouwels PJ, Barkhof F ,
Bö L. Spinal cord grey matter lesions in multiple sclerosis detected by post-mortem
high field MR imaging. Multiple Sclerosis Journal. 2009 Feb;15(2):180-8.

Gottesman II, Gould TD (2003) The endophenotype concept in psychiatry: etymology
and strategic intentions. Am J Psychiatry; 160(4):63645.

Hassan-Smith G, Douglas MR. Epidemiology and diagnosis of multiple sclerosis. British
journal of hospital medicine (London, England: 2005). 2011 Oct;72(10):M146-51.

Hill WG, Goddard ME, Visscher PM. Data and theory point to mainly additive genetic
variance for complex traits. PLoS genetics. 2008 Feb 29;4(2):e1000008.

Inglese M, Bester M. Diffusion imaging in multiple sclerosis: research and clinical
implications. NMR in Biomedicine. 2010 Aug 1;23(7):865-72.

Ioannidis JP , Ntzani EE, Trikalinos TA, Contopoulos Ioannidis DG. Replication validity of
genetic association studies. Nat Genet. 2001 Nov;29(3):3069. PubMed PMID:
11600885.

Ioannidis JP . Genetic associations: false or true? Trends Mol Med. 2003 Apr;9(4):1358.
Review. PubMed PMID: 12727138.

Ioannidis JP . Why most published research findings are false. PLoS Medicine.
2005;2(8):e124.

Ioannidis JP , Tarone R, McLaughlin JK. The false positive to falsen egative ratio in
epidemiologic studies. Epidemiology. 2011 Jul;22(4):4506. doi:
10.1097/EDE.0b013e31821b506e. PubMed PMID: 21490505.

Isobe N, Keshavan A, Gourraud PA, Zhu AH, Datta E, Schlaeger R, Caillier SJ,
Santaniello A, Lizée A, Himmelstein DS, Baranzini SE. Association of HLA genetic risk
burden with disease phenotypes in multiple sclerosis. JAMA neurology. 2016 Jul
1;73(7):795-802.

Jersild C, Svejgaard A, Fog T. HL-A antigens and multiple sclerosis. The lancet. 1972
Jun 3;299(7762):1240-1.

42

Langers DR, Jansen JF , Backes WH. Enhanced signal detection in neuroimaging by
means of regional control of the global false discovery rate. Neuroimage. 2007 Oct
15;38(1):43-56.

Li Y, Singer NG, Whitbred J, Bowen MA, Fox DA, Lin F . CD6 as a potential target for
treating multiple sclerosis. Proceedings of the National Academy of Sciences. 2017 Feb
16:201615253.

Lill CM, Schjejde BM, Graetz C, Ban M, Alcina A, Ortiz MA, Pérez J, Damotte V, Booth
D, Lopez de Lapuente A, Broer L. MANBA, CXCR5, SOX8, RPS6KB1 and ZBTB46 are
genetic risk loci for multiple sclerosis. Brain. 2013;136(6):1778-82.

Longbrake EE, Hafler DA. Linking genotype to clinical phenotype in multiple sclerosis: in
search of the holy grail. JAMA neurology. 2016 Jul 1;73(7):777-8.

Meyer-Lindenberg A, Weinberger DR. Intermediate phenotypes and genetic
mechanisms of psychiatric disorders. Nature Reviews Neuroscience. 2006 Oct
1;7(10):818-27.

Patsopoulos NA, Baranzini SE, Santaniello A, Shoostari P , Cotsapas C, Wong G,
Beecham AH, James T, Replogle J, Vlachos I, McCabe C. The Multiple Sclerosis
Genomic Map: Role of peripheral immune cells and resident microglia in susceptibility.
bioRxiv. 2017 Jan 1:143933.

Polman CH, Reingold SC, Banwell B, Clanet M, Cohen JA, Filippi M, Fujihara K,
Havrdova E, Hutchinson M, Kappos L, Lublin FD. Diagnostic criteria for multiple
sclerosis: 2010 revisions to the McDonald criteria. Annals of neurology. 2011 Feb
1;69(2):292-302.

Roshchupkin GV, Adams HH, van der Lee SJ, Vernooij MW, van Duijn CM, Uitterlinden
AG, van der Lugt A, Hofman A, Niessen WJ, Ikram MA. Fine-mapping the effects of
Alzheimer's disease risk loci on brain morphology. Neurobiology of aging. 2016 Dec
31;48:204-11.

Sawcer S, Hellenthal G, Pirinen M, Spencer CC, Patsopoulos NA, Moutsianas L, Dilthey
A, Su Z, Freeman C, Hunt SE, Edkins S. Genetic risk and a primary role for
cell-mediated immune mechanisms in multiple sclerosis. Nature. 2011 Aug
11;476(7359):214.

Swaminathan B, Cuapio A, Alloza I, Matesanz F , Alcina A, García-Barcina M, Fedetz M,
Fernández Ó, Lucas M, Órpez T, Pinto-Medel MJ. Fine mapping and functional analysis
of the multiple sclerosis risk gene CD6. PloS one. 2013 Apr 24;8(4):e62376.

43

Wagner M, Bilinska M, Pokryszko-Dragan A, Sobczynski M, Cyrul M, Kusnierczyk P ,
Jasek M. ALCAM and CD6—multiple sclerosis risk factors. Journal of
neuroimmunology. 2014 Nov 15;276(1):98-103.

Wang Y, Song Y , Rajagopalan P , An T, Liu K, Chou YY , Gutman B, Toga AW, Thompson
PM, Alzheimer's Disease Neuroimaging Initiative. Surface-based TBM boosts power to
detect disease effects on the brain: an N= 804 ADNI study. Neuroimage. 2011 Jun
15;56(4):1993-2010.

Werring DJ, Brassat D, Droogan AG, Clark CA, Symms MR, Barker GJ, MacManus DG,
Thompson AJ, Miller DH. The pathogenesis of lesions and normal-appearing white
matter changes in multiple sclerosis: a serial diffusion MRI study. Brain. 2000 Aug
1;123(8):1667-76.

Zhou X, Stephens M. Genome-wide efficient mixed-model analysis for association
studies. Nature genetics. 2012 Jul 1;44(7):821-4.


44

CHAPTER FIVE
Multiple sclerosis polygenic risk scores and white-matter integrity

5.1 Introduction

Candidate variants are targets for functional validation studies, such as the experiment
described in Chapter 4. Yet, given that a single variant on its own confers a relatively
small risk increase , methods to summarize genetic predisposition are needed.
2

The MS polygenic risk score is a useful way to encapsulate the genetic risk for MS.
Simply, the score is calculated from the sum of MS risk alleles weighted by their SNP
effect from the GWAS in which they were discovered. A similar approach has been
applied to various other complex multigenic diseases, including Alzheimer’s disease,
cardiovascular disease, stroke risk, breast cancer, and schizophrenia. In MS, patients
show significantly increased polygenic risk scores compared to controls and the scores
correlate with age of onset (De jager et al., 2009, Gourraud et al., 2011). It is strongly
associated with oligoclonal banding, a cerebrospinal fluid (CSF) biomarker of MS
(Harbo et al., 2013) and Non-HLA based MS risk scores predict relapse rates (Hilven et
al., 2015). It is, however, not currently known how MS genetic risk is expressed in MRI.

Here, we test the effect of MS polygenic risk scores on white-matter in DTI in healthy
young adults. Employing a similar approach to the methods in Chapter 4, we first
explore the effect of MS PR scores on FA, building upon our single allele experiments.
We also overlaid our statistical result maps with whole-brain gene expression data from
the Allen Brain Atlas of the HLA-DRB1* allele, which is likely a driver of the PR score.

Lesion location is highly heterogeneous across MS patients, and it is unclear whether
white-matter damage occurs predictably in overlapping regions across subjects
(Gilmore et al., 2009). Traditionally, statistical MRI analysis methods look for consistent
regions in which there is an effect. As there isn’t necessarily a reason to assume global
genetic risk for MS will affect the same brain regions across subjects, we employ a
method known as white-matter ‘potholes,’ which searches for within-subject areas of
aberrant FA. This allows us to examine the effect of MS genetic risk on white-matter
without assuming spatial overlap across subjects.

2
Chapters 1 and 4 describe single-SNP association limitations in detail
45

5.2 Methods

5.2.1 Subjects and genotyping

See Chapter 4 for a detailed description of the subject sample, genotyping methods
and imaging parameters.

5.2.2 Polygenic risk scores

Polygenic risk scores (PRS) were generated for each subject using PLINK as described
by Purcell et al. First we performed LD pruning on a list of the top 102 SNPs listed in
Supplementary Table A of the IMSGC GWAS. The resulting 76 SNPs were used to
estimate individual polygenic risk scores, based on number of risk alleles an individual
possessed weighted by the standardized SNP effect observed in the original GWAS.
Final scores were normalized and account for any missing SNPs.

5.2.3 Risk score association

Associations with scores and FA were tested using mixed-model regression at each
white matter voxel in FA maps, which controlled for age, sex and accounted for kinship.
Langers Searchlight FDR was used to control for multiple comparisons (Langers,
Jansen, and Backes, 2007).

5.2.4 White-matter potholes

White-matter “potholes” were quantified for each subject by first creating Z-
transformed FA maps, where a pothole was defined as any voxel with a preset number
of standard deviations away from the subject’s mean (at different cutoffs on the range
.5-3). A more detailed description of the pothole method can be found in White et al.
2009. Again, linear regression was used to relate potholes to polygenic risk scores,
controlling for age and sex.

5.2.5 Gene expression maps

The HLA-DRB1* Allele confers the strongest known odds ratio for MS risk, and likely is
a strong driver of the PRS signal. To test this, we employed the same method described
in Chapter 4.2.6. Here we overlaid gene expression maps of HLA-DRB1* from the Allen
Brain Atlas with p-value result maps from the analysis described in 5.2.3.

46

5.3 Results

Using a similar approach as the methods detailed in Chapter 4.2, we tested the effect of
MS polygenic risk scores on FA maps in 398 healthy young adults. Greater polygenic
risk for developing MS was significantly (FDR critical q value = 0.05) higher FA in several
regions throughout the brain, including the Corpus Callosum, L Inferior longitudinal
fasciculus, and R Anterior Thalamic radiation.

Figure 5.1: Statistical maps generated from 398 young adults are shown below. FA in several regions was significantly
associated (q < 0.05) with polygenic risk for MS, including Corpus Callosum, L Inferior longitudinal fasciculus, and R
Anterior Thalamic radiation. Each colored region was significant at FDR critical q value. Colors indicate beta values.

In 3/5 donors HLA-DRB1* was significantly overexpressed in voxels associated with the
MS PRs (min p=.008).

Figure 5.2: Overlay of Allen Brain Atlas HLA-DRB1* gene expression maps with results from white-matter association
study. Colored regions are significant at FDR critical value. Color bars below images show gene expression z-score by
brain region.
47

Figure 5.3 MS Polygenic risk score by number of white-matter potholes at 2 standard deviations. The x-axis shows voxel
counts when potholes are defined as any voxel with 2 or greater standard deviations away from the average. Y-axis is
normalized MS-PRS. Results were significant at (P=.03).

48

Figure 5.4 MS Polygenic risk score by number of white-matter potholes at .5 standard. The x-axis shows voxel counts
when potholes are defined as any voxel with 2 or greater standard deviations away from the average. Y-axis is
normalized MS-PRS. Results here failed to reach significance (P=.08).


49

5.4 Discussion

Moving beyond single-SNP associations, we found that higher MS polygenic risk
scores were associated with higher FA in several brain regions, including the
white-matter tracts Corpus Callosum, left Inferior longitudinal fasciculus, and right
anterior thalamic radiation. Polygenic risk scores are an attempt to summarize the total
genetic risk of a polygenic disease for a given subject.

We also found regions in which the HLA-DRB1* allele was significantly overexpressed in
regions which overlapped with our statistical result maps of PR score and white-matter
association. It is possible that the HLA allele is driving the signal, or is a necessary
component of MS risk.

DTI studies of MS usually report reduced FA in MS patients, both in lesions and NAWM
(Longoni et al., 2017, Cercignani et al., 2001; Roosendall et al., 2009), although there
are reported findings of regions with higher FA. Here we found higher FA with increased
MS genetic risk. While higher FA is generally interpreted as increased myelination,
better fiber coherence and improved function, it is not a direct measure of WM and can
be explained be several factors such as axonal deficits (known to occur in early stages
of the disease [Trapp et al., 1998]), decreases in axonal diameter, packing density, and
branching (Beaulieu, 2002). It can also be explained by imaging factors such as fewer
intravoxel crossing fibers (Hoeft et al., 2007). FA has also been shown to be higher in
neurogenetic diseases coupled with aberrant function (Braskie et al., 2011).

Similar to our findings in reported in Chapter 4, in this study, we found clusters of higher
FA—either as overlapping clusters or potholes—can be pre-lesional sites of increased
risk that either flag or attract demyelinating factors to act.

Finally, we also found a positive association between MS PRS and white-matter
potholes, which are spatially independent clusters of aberrant FA across subjects. We
found that number of potholes increased with MS PR score, again indicating that MS
genetic risk modulates white matter integrity in healthy individuals.

In sum, these findings demonstrate clear evidence that genetic predisposition for MS
affects white matter (WM) in healthy participants with no documented symptoms of the
disease.

50

5.5 Chapter 5 References

Abdollah Zadeh R, Jalilian N, Sahraian MA, Kasraian Z, Noori-Daloii MR.
Polymorphisms of RPS6K B1 and CD86 associates with susceptibility to multiple
sclerosis in Iranian population. Neurological research. 2017 Mar 4;39(3):217-22.

Ascherio A, Munger KL, White R, Köchert K, Simon KC, Polman CH, Freedman MS,
Hartung HP , Miller DH, Montalbán X, Edan G. Vitamin D as an early predictor of multiple
sclerosis activity and progression. JAMA neurology. 2014 Mar 1;71(3):306-14.

Baranzini SE, Oksenberg JR. The Genetics of Multiple Sclerosis: From 0 to 200 in 50
Years. Trends in Genetics. 2017 Oct 5.

Cercignani M, Inglese M, Pagani E, Comi G, Filippi M. Mean diffusivity and fractional
anisotropy histograms of patients with multiple sclerosis. American Journal of
Neuroradiology. 2001 May 1;22(5):952-8.

D’Cunha MA, Pandit L, Malli C. CD6 gene polymorphism rs17824933 is associated with
multiple sclerosis in Indian population. Annals of Indian Academy of Neurology. 2016
Oct;19(4):491.

Da Gloria, V.G., Martins De Araujo, M., Mafalda Santos, A., Leal, R., De Almeida, S.F .,
Carmo, A.M., Moreira, A., 2014. T cell activation regulates CD6 alternative splicing by
transcription dynamics and SRSF1. J. Immunol. 193, 391–399.

De Jager PL, Chibnik LB, Cui J, Reischl J, Lehr S, Simon KC, Aubin C, Bauer D,
Heubach JF , Sandbrink R, Tyblova M. Integration of genetic risk factors into a clinical
algorithm for multiple sclerosis susceptibility: a weighted genetic risk score. The Lancet
Neurology. 2009 Dec 31;8(12):1111-9.

Dobson R, Ramagopalan S, Topping J, Smith P , Solanky B, Schmierer K, Chard D,
Giovannoni G. A Risk Score for Predicting Multiple Sclerosis. PloS one. 2016 Nov
1;11(11):e0164992.

Fedetz M, Fernández O, Lucas M, Orpez T, Pinto-Medel MJ, Otaegui D, Olascoaga J,
Gilmore CP , Donaldson I, Bö L, Owens T, Lowe J, Evangelou N. Regional variations in
the extent and pattern of grey matter demyelination in multiple sclerosis: a comparison
between the cerebral cortex, cerebellar cortex, deep grey matter nuclei and the spinal
cord. Journal of Neurology, Neurosurgery & Psychiatry. 2009 Feb 1;80(2):182-7.

Gilmore CP , Geurts JJ, Evangelou N, Bot JC, Van Schijndel RA, Pouwels PJ, Barkhof F ,
Bö L. Spinal cord grey matter lesions in multiple sclerosis detected by post-mortem
high field MR imaging. Multiple Sclerosis Journal. 2009 Feb;15(2):180-8.
51

Giulia Longoni, Robert A Brown, Parya Momayyez Siahkal, Colm Elliott, Sridar
Narayanan, Amit Bar-Or, Ruth Ann Marrie, E Ann Yeh, Massimo Filippi, Brenda Banwell,
Douglas L Arnold, on behalf of The Canadian Pediatric Demyelinating Disease Network;
White matter changes in paediatric multiple sclerosis and monophasic demyelinating
disorders, Brain, Volume 140, Issue 5, 1 May 2017, Pages 1300–1315,

Gourraud PA, McElroy JP , Caillier SJ, Johnson BA, Santaniello A, Hauser SL, Oksenberg
JR. Aggregation of multiple sclerosis genetic risk variants in multiple and single case
families. Annals of neurology. 2011 Jan 1;69(1):65-74.

Harbo HF , Isobe N, Berg-Hansen P , Bos SD, Caillier SJ, Gustavsen MW, Mero IL, Celius
EG, Hauser SL, Oksenberg JR, Gourraud PA. Oligoclonal bands and age at onset
correlate with genetic risk score in multiple sclerosis. Multiple Sclerosis Journal. 2014
May;20(6):660-8.

Hassan-Smith G, Douglas MR. Epidemiology and diagnosis of multiple sclerosis. Br J
Hosp Med (Lond). 2011 Oct;72(10):M14651. Review. PubMed PMID: 22041658.

Hilven K, Patsopoulos NA, Dubois B, Goris A. Burden of risk variants correlates with
phenotype of multiple sclerosis. Multiple Sclerosis Journal. 2015 Nov;21(13):1670-80.

Hoeft F , Barnea-Goraly N, Haas BW, Golarai G, Ng D, Mills D, Korenberg J, Bellugi U,
Galaburda A, Reiss AL. More is not always better: increased fractional anisotropy of
superior longitudinal fasciculus associated with poor visuospatial abilities in Williams
syndrome. Journal of Neuroscience. 2007 Oct 31;27(44):11960-5.

Inglese M, Bester M. Diffusion imaging in multiple sclerosis: research and clinical
implications. NMR in Biomedicine. 2010 Aug 1;23(7):865-72.

International Multiple Sclerosis Genetics Consortium. Evidence for polygenic
susceptibility to multiple sclerosis—the shape of things to come. The American Journal
of Human Genetics. 2010 Apr 9;86(4):621-5.

Langers DR, Jansen JF , Backes WH. Enhanced signal detection in neuroimaging by
means of regional control of the global false discovery rate. Neuroimage. 2007 Oct
15;38(1):43-56.

Li Y, Singer NG, Whitbred J, Bowen MA, Fox DA, Lin F . CD6 as a potential target for
treating multiple sclerosis. Proceedings of the National Academy of Sciences. 2017 Feb
16:201615253.

52

Patsopoulos NA, Baranzini SE, Santaniello A, Shoostari P , Cotsapas C, Wong G,
Beecham AH, James T, Replogle J, Vlachos I, McCabe C. The Multiple Sclerosis
Genomic Map: Role of peripheral immune cells and resident microglia in susceptibility.
bioRxiv. 2017 Jan 1:14393

PLoS One. 2013 Apr 24;8(4):e62376. doi: 10.1371/journal.pone.0062376. Print 2013.
PubMed PMID: 23638056; PubMed Central PMCID: PMC3634811.
Sawcer S, Hellenthal G, Pirinen M, Spencer CC, Patsopoulos NA, Moutsianas L, Dilthey
A, Su Z, Freeman C, Hunt SE, Edkins S. Genetic risk and a primary role for
cell-mediated immune mechanisms in multiple sclerosis. Nature. 2011 Aug
11;476(7359):214.

Swaminathan B, Cuapio A, Alloza I, Matesanz F , Alcina A, García-Barcina M,
Urcelay E, Ortiz MA, Arroyo R, Oksenberg JR, Antigüedad A, Tolosa E, Vandenbroeck
Werring DJ, Brassat D, Droogan AG, Clark CA, Symms MR, Barker GJ, MacManus DG,
Thompson AJ, Miller DH. The pathogenesis of lesions and normal-appearing white
matter changes in multiple sclerosis: a serial diffusion MRI study. Brain. 2000 Aug
1;123(8):1667-76.

Swaminathan B, Cuapio A, Alloza I, Matesanz F , Alcina A, García-Barcina M, Fedetz M,
Fernández Ó, Lucas M, Órpez T, Pinto-Medel MJ. Fine mapping and functional analysis
of the multiple sclerosis risk gene CD6. PloS one. 2013 Apr 24;8(4):e62376.

Trapp BD, Peterson J, Ransohoff RM, Rudick R, Mörk S, Bö L. Axonal transection in the
lesions of multiple sclerosis. New England Journal of Medicine. 1998 Jan
29;338(5):278-85.

Zhou X, Stephens M. Genome-wide efficient mixed-model analysis for association
studies. Nature genetics. 2012 Jul 1;44(7):821-4.


53

CHAPTER SIX
Genetic connectivity: correlated genetic control of
cortical thickness, brain volume and white-matter

6.1 Introduction

Understanding the degree to which different brain measures are affected by shared
genetic influences is valuable in explaining brain changes during development and
emerging pathology; it can also help experimental design in the field of imaging
genetics (Winkler et al., 2009). Brain structures are under strong genetic control
(Blokland et al., 2012), but patterns of common genetic influence across multiple brain
measures (and across different imaging modalities) are still largely unknown.
Understanding the genetic relationship between different brain measures may offer
biologically meaningful targets for genetic analysis, revealing how they are
interconnected. In searching for genes that influence brain measures, it would be
advantageous to select imaging measures that are genetically distinctive or
unique—discovering genes that help shape multiple brain regions is also of great
interest.

Twin and family studies can estimate the degree of common genetic influence
underlying any two traits (called the genetic correlation or r
g
). By clustering cortical
regions that are genetically correlated, Fjell et al. recently noted modules or sectors of
the cortex that appear to develop and age in distinctive ways (Fjell et al., 2015;
Thompson, 2015). Panizzon et al. found that cortical surface area and thickness are
influenced by separate genetic factors—they suggested that surface or thickness
measures may be better targets for gene discovery studies than cortical volume
measures, which contain genetic and phenotypic aspects of both (Panizzon et al.,
2009). A corollary of this work is to examine measure across modalities and possibly to
develop a trans- modality measure for gene discovery.

Diffusion tensor imaging (DTI) is commonly acquired along with standard T1-weighted
structural magnetic resonance imaging (MRI) scans and has the potential to describe
white matter (WM) integrity across the entire brain. The relationship between DTI
metrics and brain morphometry is largely unstudied. WM tracts imaged in DTI
interconnect many subcortical regions commonly studied in imaging genetics. They
also develop in concert with cortical and subcortical gray matter as neural pruning and
fiber organization occurs (Casey et al., 2005). The genetic control of this development is
not yet understood.

Here we examined 442 healthy, young adult twins to estimate the genetic correlation
between measures from 109 brain regions of interest (ROIs; 21 measures of WM
fractional anisotropy (FA), 15 of subcortical volume, 72 of cortical thickness, and the
intracranial volume). These measures were computed with standardized protocols
54

developed by the Enhancing Neuroimaging Genetics through Meta-Analysis (ENIGMA)
Consortium. The ENIGMA Consortium is a large multisite imaging genetics collaboration
involving over 185 laboratories across 35 countries. ENIGMA recently published a
genome-wide association study of subcortical volumes, using MRI and genotyping data
from 30,717 individuals (Hibar et al., 2015); similar studies of DTI and cortical measures
are now underway.

Here we evaluated the genetic relationships among WM, subcortical volumetrics, and
cortical thickness and surface area. Given the developmental and physical links
between WM and cortical thickness, surface area, and subcortical volumetrics, we
hypothesized that the underlying genetic determinants of each of those measures
would show some overlap with those measuring WM integrity.

6.2 Methods

Bivariate genetic correlations were computed between ROIs for three imaging
measures: subcortical volume, cortical thickness, and DTI FA, as described in the
ENIGMA protocols.

6.2.1 Subject information

A total of 442 subjects (mean age: 23.5 years 2.1 SD; 151 women; 98 monozygotic
(MZ) pairs, 123 dizygotic (DZ) pairs) were included; all subjects underwent structural
T1-weighted brain MRI and DTI scans. All subjects were of European ancestry from 221
families. Subjects were recruited as part of a large-scale 5-year twin study examining
healthy young adult Australians using structural and functional MRI and DTI (Zubicaray
et al., 2008).

6.2.2 Image Acquisition

Structural and diffusion-weighted (DW) whole-brain MRI scans were acquired for every
participant (on a 4T Bruker Medspec scanner). T1-weighted images were acquired with
an inversion recovery rapid gradient echo sequence (TI/TR/TE 1⁄4 700/1500/3.35 ms;
flip angle 1⁄4 8 degrees; slice thickness 1⁄4 0.9 mm, with a 2563 acquisition matrix).

DW images were acquired using single-shot echo planar imaging with a
twice-refocused spin-echo sequence to reduce eddy-current-induced distortions. A
3-min, 30- gradient acquisition was designed to optimize signal-to- noise ratio for
diffusion tensor estimation (58). Imaging parameters were: TR/TE 1⁄4 6090/91.7 ms,
FOV 1⁄4 23 cm, with a 128 128 acquisition matrix. Each 3D volume consisted of 55
2-mm thick axial slices and 1.8 1.8 mm2 in-plane resolution. One hundred and five
images were acquired per subject: 11 with no diffusion sensitization (i.e., T2-weighted
55

b0 images) and 94 DW images (b 1⁄4 1149 s/mm2) with gradient directions uniformly
distributed on the hemisphere.

6.2.3 Image Preprocessing

All images were processed as described by the publicly available ENIGMA image
analysis protocols (http://enigma.ini. usc.edu/protocols/imaging-protocols/).

6.2.4 Establishing Zygosity, Genotyping, and Imputation

Standard polymerase chain reaction (PCR) methods and genotyping were used to
establish zygosity objectively by typing nine independent DNA microsatellite
polymorphisms (polymorphism information content >0.7). Blood group (ABO, MNS, and
Rh) was used to verify results along with phenotypic data (hair, skin, and eye color),
providing a probability of accurate zygosity classification >99.99%. Standard
manufacturer protocols were used on the Human610-Quad BeadChip (Illumina) to
analyze genomic DNA samples (Infinium HD Assay; Super Protocol Guide; Rev. A, May
2008). Genotypes were imputed by mapping to HapMap (Release 22, Build 36) with
MACH (http://www.sph.umich.edu/csg/abecasis/ MACH/index.html).

6.2.5 Cross-Twin Cross-Trait Analysis

To identify common genetic or environmental factors modulating cortical thickness,
subcortical volume, and DTI FA measures, we used a “cross-twin cross-trait” analysis
(Neale et al., 1992). Covariance matrices for the MRI and DTI measures were computed
between the MZ twins who share all the same genes, and the DZ twins who on average
share half of their genetic polymorphisms.

Using OpenMx software (http://openmx.psyc.virginia. edu/), covariance matrices were
entered into a multivariate structural equation model (SEM) to estimate the relative
contributions of additive genetic (A), shared environmental (C), and unique
environmental (E) components to the population variances and covariances of the
observed variables. The E component also contains experimental measurement error
and is assumed to be independent between both twins in a pair.

In multivariate SEM, it is expected that there are common genetic and environmental
factors affecting various brain measures. We can estimate the variance of the common
genetic and environmental components from the total population variance by
calculating the difference in the co- variances between the MZ and DZ twins within the
same individual (cross-trait within individual) and also between one phenotype in one
twin with the other phenotype in the second twin (cross-twin cross-trait). With this
multivariate SEM, we obtain ra and rc, which denote the additive genetic and shared
environmental influences on the correlations between the two phenotypes, respectively.
56

The cross-trait within individual correlation (the correlation between two brain measures,
FA and volume for example, in twin 1 or in twin 2) can be split into the additive genetic
and shared and unique environmental components (e.g., A
V,i
, C
V,i
, and E
V,i
for each ROI
value), and the correlation coefficients between A
V,i
and A
FA,i
, C
V,i
and C
FA,i
, and E
V,i
and
E
FA,i
, are indicated by r
a
, r
c
, and r
e
, respectively. The cross-trait cross-twin correlation is
shown as A
V,i
and A
FA,j
, and C
V,i
and C
FA,j
for the FA value in twin i and the volume value in
twin j, where i, j = 1 or 2, and i ≠j. Because the unique environmental factors between
subjects is considered to be independent, there is no r
e
term for E
V,i
and E
T,j
.

Using the path diagram, we derive the covariance across the two phenotypes within the
same subject (or separately in the two subjects) by multiplication of the path
coefficients for the closed paths (Fig. 6.1).

Figure 6.1 Example structural equation model path diagram for bivariate association. FA, fractional anisotropy.

For example, covariance between the FA values in twin 1 and the volume in twin 2 is
equal to a
V
·r
a
·a
T
+c
V
·r
c
·c
T
for MZ twins, and a
V
·½r
a
·a
T
+c
V
·r
c
·c
T
for DZ twins. Paths
connecting the same phenotype are identical to a univariate SEM model (Zubicaray et
al., 2008). MZ twins have a correlation coefficient of 1 for A1 and A2, while DZ twins
have 0.5. By definition for the shared environment, C1 and C2 is always a correlation
coefficient of 1. E1 and E2 are assumed to have no correlation.

57

It is common in twin studies to test whether the observed measures are best modeled
using a combination of additive genetic, shared, and unshared environmental factors, or
whether only one or two of these factors is sufficient to explain the observed pattern of
inter-twin correlations. More details on model selection are described in Jahanshad et
al. (2012). Here we used the full set of path coefficients in each test as they achieved
significance.

6.2.6 Phenotypic correlations

To assess the phenotypic relationship across measures, correlation matrices were
generated using cor.test as a part of the R software package
(https://www.r-project.org/), across all measurements entered into the genetic analysis.

6.2.7 Multiple Comparisons Correction

To control for multiple comparisons, the standard Benjamini & Hochberg FDR procedure
(1995) was employed in all statistical tests: in determining the best overall model for the
SEM cross-twin cross-trait analysis, and in the phenotypic and genetic correlation
analysis (q = 0.05).

6.3 Results

Bivariate genetic correlations between subcortical volume and FA measures are shown
in Fig. 6.2. For tests of genetic correlation between cortical thickness and FA, the only
significant correlations were of left and right hemisphere surface area, intracranial
volume (ICV), and various DTI tracts listed below in Fig. 6.3. Similarly in tests of genetic
correlation between cortical thickness and volume, only surface area and ICV were
significant, listed in Fig. 6.4. The results of the phenotypic correlations between FA and
volume are shown in Fig. 6.5. Phenotypic correlations between cortical thickness and
volume are shown in Fig. 6.6.

In the phenotypic correlation test between thickness and FA the only significant findings
were between the genu of the corpus callosum (GCC) and left surface area, right
surface area and ICV (r = .22, .23, .22, respectively, q < .05); and the anterior corona
radiata (ACR) and left surface area and right surface area (r = .23, .23, respectively, q <
.05).

58

Figure 6.2 Genetic correlations between fractional anisotropy in diffusion tensor imaging (DTI) white matter tracts and
subcortical volume regions of interest. Colored elements (light gray in print versions) indicate significant associations
after false discovery rate correction. Warmer colors (darker colors in print versions) indicate higher genetic correlation (rg)
values. Some very small structures (e.g., the nucleus accumbens) did not show significant correlations with any of the
DTI measures. Accumb, nucleus accumbens; ACR, anterior corona radiata; ALIC, anterior limb of the internal capsule;
Amyg, amygdala; BCC, body of the corpus callosum; Caud, caudate; CC, corpus callosum; CGC, cingulum; CHG,
hippocampal part of the cingulum; CR, corona radiata; EC, external capsule; FXST, fornix/stria terminalis; GCC, genu of
the corpus callosum; Hippo, hippocampus; IC, internal capsule; ICV, intracranial volume; Pal, globus pallidus; PCR,
posterior corona radiata; PLIC, posterior limb of the internal capsule; PTR, posterior thalamic radiation; Put, putamen;
RLIC, retrolenticular part of the internal capsule; SCC, subcallosal cingulate white matter; SCR, su- perior region of
corona radiata; SFO, superior fronto-occipital fasciculus; SLF , superior longitudinal fasciculus; SS, sagittal striatum; Thal,
thalamus.

59

Figure 6.3 Genetic correlations between fractional anisotropy in diffusion tensor imaging white matter tracts, cortical
surface area, and intracranial volume (ICV). Colored elements (light gray in print versions) indicate significant associations
after false discov-ery rate correction. Warmer colors (darker colors in print versions) indicate higher genetic correlation
(rg) values. ACR, anterior corona radiata; ALIC, anterior limb of the internal capsule; BCC, body of the corpus callosum;
CC, corpus callosum; CGC, cingulum; CHG, hippocampal part of the cingulum; CR, corona radiata; EC, external
capsule; FXST, fornix/stria terminalis; GCC, genu of the corpus callosum; IC, internal capsule; PCR, posterior corona
radiata; PLIC, posterior limb of the internal capsule; PTR, posterior thalamic radiation; RLIC, retrolenticular part of the
internal capsule; SCC, subcallosal cingulate white matter; SCR, superior region of corona radiata; SFO, superior
fronto-occipital fasciculus; SLF , superior longitudinal fasciculus; SS, sagittal striatum; SurfArea, cortical surface area.

60

Figure 6.4 Genetic correlations between subcortical volume regions of interest and cortical surface area measures.
Colored elements (light gray in print versions) indicate significant associations after false discovery rate correction.
Warmer colors (darker colors in print versions) indicate higher genetic correlation (rg) values. Accumb, nucleus
accumbens; Amyg, amygdala; Caud, caudate; Hippo, hippocampus; ICV, intracranial volume; Pal, globus pallidus; Put,
putamen; SurfArea, cortical surface area; Thal, thalamus.


61



Figure 2.5 Phenotypic correlations between fractional anisotropy and volume measures. Colored elements (light gray in
print versions) indicate significant associations after false discovery rate correction. Warmer colors (darker colors in print
versions) indicate higher correlation (Pearson’s r) values. Accumb, nucleus accumbens; ACR, anterior corona radiata;
ALIC, anterior limb of the internal capsule; Amyg, amygdala; BCC, body of the corpus callosum; Caud, caudate; CC,
corpus callosum; CGC, cingulum; CHG, hippocampal part of the cingulum; CR, corona radiata; EC, external capsule;
FXST, fornix/stria terminalis; GCC, genu of the corpus callosum; Hippo, hippocampus; IC, internal capsule; ICV,
intracranial volume; Pal, globus pallidus; PCR, posterior corona radiata; PLIC, posterior limb of the internal capsule; PTR,
posterior thalamic radiation; Put, putamen; RLIC, retrolenticular part of the internal capsule; SCC, subcallosal cingulate
white matter; SCR, su- perior region of corona radiata; SFO, superior fronto-occipital fasciculus; SLF , superior
longitudinal fasciculus; SS, sagittal striatum; SurfArea, cortical surface area; UNC, uncinate fasciculus; Thal, thalamus.


62

Figure 6.6 Phenotypic correlations between cortical thickness, surface area, and volume measures. Colored elements
(light gray in print versions) indicate significant associations after FDR correction. Warmer colors (darker colors in print
versions) indicate higher correlation (Pearson’s r) values. Accumb, nucleus accumbens; Amyg, amygdala; Caud, caudate;
Hippo, hippocampus; ICV, intracranial volume; Pal, globus pallidus; Put, putamen; SurfArea, cortical surface area; Thal,
thalamus.


63

6.4 Discussion

We used cross-trait structural equation modeling in a twin design to study the common
genetic influences among three categories of common structural MRI and DTI brain
measures: cortical thickness, subcortical volumes, and FA in WM tracts. We found
significant genetic correlations between FA in several WM tracts and subcortical volume
ROIs, notably the thalamus and pallidum. Association between cortical thickness and
both volume ROIs and WM was not detectable, even in this relatively large sample.

Even so, cortical surface area for each hemisphere was highly correlated with FA in
several WM tracts and all of the subcortical volume measures. Prior studies have
suggested correlated genetic control across different brain structures (Glahn,
Thompson and Blangero 2007). Our finding of moderate genetic and phenotypic
overlap between FA in several WM tracts and subcortical volumes is not surprising. Our
strongest associations included the thalamus, pallidum, and several other basal ganglia
structures. These structures are highly networked by WM tracts in cortico-basal
ganglia-thalamocortical loops. Furthermore, they are physically closer to each other
than to cortical gray matter.

Measures of thickness from the cortex—in addition to being structurally further away
from subcortical regions—may also be affected by genetic programs that arose more
recently in evolution. Cortical thickness is influenced largely by the number of neurons
in a cortical column, according to the radial unit hypothesis of cortical development
(Rakick 1995). Surface area and thickness have different genetic origins (Winkler et al.,
2009), and MRI measurements of cortical thickness are not strongly associated with ICV
or other head size measures (Im et al., 1991). This could partially explain the lack of
association between cortical thickness and the other measures examined here.

ICV was genetically correlated with all subcortical volume measures, albeit modestly,
with values ranging from 0.1 to 0.29. The phenotypic correlation between these
measures was greater, ranging from 0.29 to 0.65. This may indicate that the relationship
between these measures is not completely driven by common genetic factors.

There was some agreement between the genetic and phenotypic correlations, with the
strongest phenotypic results being between FA in the left and right thalamus and the
GCC. We did not detect an association between most of thickness measurements and
any of the other phenotypes. In the future it may be useful to cluster regions that show
genetic correlations (or even phenotypic correlations, as they tend to be quite similar).

Our primary aim was to identify genetically correlated—and conversely
independent—regions across derived MRI measures that are commonly used in
imaging genetics studies. These results can inform future gene discovery efforts, and
clustering of these measures may boost power and allow dimensionality reduction.
64

These correlation maps may also help elucidate the genetic control of the anatomy
underlying common imaging measures.

Here we examined young adult subjects of European ancestry. Future studies of other
populations would be valuable to test whether these genetic associations hold across
different ethnicities.


65

6.5 Chapter 6 References

Benjamini Y , Hochberg Y. Controlling the false discovery rate: a practical and powerful
approach to multiple testing. Journal of the royal statistical society. Series B
(Methodological). 1995 Jan 1:289-300.

Blokland GA, de Zubicaray GI, McMahon KL, Wright MJ. Genetic and environmental
influences on neuroimaging phenotypes: a meta-analytical perspective on twin imaging
studies. Twin Research and Human Genetics. 2012 Jun;15(3):351-71.

Casey BJ, Tottenham N, Liston C, Durston S. Imaging the developing brain: what have
we learned about cognitive development?. Trends in cognitive sciences. 2005 Mar
31;9(3):104-10.

De Zubicaray GI, Chiang MC, McMahon KL, Shattuck DW, Toga AW, Martin NG, Wright
MJ, Thompson PM. Meeting the challenges of neuroimaging genetics. Brain imaging
and behavior. 2008 Dec 1;2(4):258.

Fjell AM, Grydeland H, Krogsrud SK, Amlien I, Rohani DA, Ferschmann L, Storsve AB,
Tamnes CK, Sala-Llonch R, Due-Tønnessen P , Bjørnerud A. Development and aging of
cortical thickness correspond to genetic organization patterns.

Proceedings of the National Academy of Sciences. 2015 Dec 15;112(50):15462-7.
Glahn DC, Thompson PM, Blangero J. Neuroimaging endophenotypes: strategies for
finding genes influencing brain structure and function. Human brain mapping. 2007 Jun
1;28(6):488-501.

Hibar DP , Stein JL, Renteria ME, Arias-Vasquez A, Desrivières S, Jahanshad N, Toro R,
Wittfeld K, Abramovic L, Andersson M, Aribisala BS. Common genetic variants
influence human subcortical brain structures. Nature. 2015 Apr 9;520(7546):224-9.

Im K, Lee JM, Lyttelton O, Kim SH, Evans AC, Kim SI. Brain size and cortical structure
in the adult human brain. Cerebral Cortex. 2008 Jan 29;18(9):2181-91.

Jahanshad N, Kohannim O, Hibar DP , Stein JL, McMahon KL, de Zubicaray GI,
Medland SE, Montgomery GW, Whitfield JB, Martin NG, Wright MJ. Brain structure in
healthy adults is related to serum transferrin and the H63D polymorphism in the HFE
gene. Proceedings of the National Academy of Sciences. 2012 Apr 3;109(14):E851-9.

Jahanshad N, Lee AD, Barysheva M, McMahon KL, de Zubicaray GI, Martin NG, Wright
MJ, Toga AW, Thompson PM. Genetic influences on brain asymmetry: a DTI study of
374 twins and siblings. Neuroimage. 2010 Aug 15;52(2):455-69.
Neale MC, Cardon LR. Methodology for genetic studies of twins and families. Springer
66

Science & Business Media; 2013 Mar 9.

Panizzon MS, Fennema-Notestine C, Eyler LT, Jernigan TL, Prom-Wormley E, Neale M,
Jacobson K, Lyons MJ, Grant MD, Franz CE, Xian H. Distinct genetic influences on
cortical surface area and cortical thickness. Cerebral cortex. 2009 Mar
18;19(11):2728-35.

Rakic P . A small step for the cell, a giant leap for mankind: a hypothesis of neocortical
expansion during evolution. Trends in neurosciences. 1995 Sep 30;18(9):383-8.

Schmitt JE, Eyler LT, Giedd JN, Kremen WS, Kendler KS, Neale MC. Review of twin and
family studies on neuroanatomic phenotypes and typical neurodevelopment. Twin
Research and Human Genetics. 2007 Oct;10(5):683-94.

Thompson PM. Cracking the brain’s genetic code. Proceedings of the National
Academy of Sciences. 2015 Dec 15;112(50):15269-70.

Winkler AM, Kochunov P , Blangero J, Almasy L, Zilles K, Fox PT, Duggirala R, Glahn
DC. Cortical thickness or grey matter volume? The importance of selecting the
phenotype for imaging genetics studies. Neuroimage. 2010 Nov 15;53(3):1135-46.

67

CHAPTER SEVEN
Future directions

7.0 Further tests of genetic pleiotropy

Publicly available GWAS summary statistics will also to further test the genetic overlap
of MS with other traits / diseases using either CIA (Chapter 2) or SECA (Chapter 3).
Particularly we plan to test overlap with Rheumatoid arthritis, which is known as the
“Bermuda triangle of genetics, environment and autoimmunity,” where like MS
HLA-DRB1 alleles are heavily implicated, and the disease is 60% heritable indicating a
complex genetic effect (Kurkó et al., 2015). A GWAS of white matter integrity is also
underway as part of the ENIGMA-DTI consortium, which will be a very useful to
compare with MS.

7.1 Polygenic risk score

Polygenic risk scores can be calculated in a number of different ways, and researchers
are now using a large number of SNPs at different P-value cutoffs as the genome-wide
significant cutoff may be too stringent. Additionally, Bayesian methods have been
shown to improve prediction (So, and Sham, 2017). Logically, prediction ability should
improve with a fine-tuning of the SNP weights, as it is unlikely that altered function is
not conferred equally by all risk variants. In MS, Dobson et al., (2015) describe a genetic
risk score that incorporates environmental factors as well.

7.2 Studies with MS patients

We plan to extend our genetic association findings to MS patients. Oslo University
Hospital is currently conducting multiple serial MRI studies in MS patients. In
collaboration with that institution we will test the effect of MS polygenic risk scores in
230 Norwegian MS patients (60% F).

7.3 Microstructural diffusion MRI

Diffusion tensor imaging has been applied to study MS lesions, but has many
documented limitations. The simple surrogate markers provided by DTI (FA, MD, ADC),
do not inform us about neurite density, neurite orientation distribution, in addition to a
number of other microstructural changes (Beaulieu, 2002; Pierpaoili and Basser, 1996).
Moreover, the hallmark complaint of DTI is its inability to capture multiple fibers, relying
only one principal direction. Advanced diffusion methods will be required going forward
to gain an understanding of the microstructural impact of the disease. As advanced
diffusion images are being acquired of MS patients in Norway as described in Section
5.1, we aim to apply our genetic association approaches with those scans.

68

7.2 Chapter 7 References

Beaulieu C. The basis of anisotropic water diffusion in the nervous system–a technical
review. NMR in Biomedicine. 2002 Nov 1;15(7‐8):435-55.

Dobson R, Ramagopalan S, Topping J, Smith P , Solanky B, Schmierer K, Chard D,
Giovannoni G. A Risk Score for Predicting Multiple Sclerosis. PloS one. 2016 Nov
1;11(11):e0164992.

Kurkó J, Besenyei T, Laki J, Glant TT, Mikecz K, Szekanecz Z. Genetics of rheumatoid
arthritis—a comprehensive review. Clinical reviews in allergy & immunology. 2013 Oct
1;45(2):170-9.

Pierpaoli C, Basser PJ. Toward a quantitative assessment of diffusion anisotropy.
Magnetic resonance in Medicine. 1996 Dec 1;36(6):893-906.

So HC, Sham PC. Improving polygenic risk prediction from summary statistics by an
empirical Bayes approach. Scientific Reports. 2017 Feb 1;7:41262.
69

Abstract (if available)

Abstract Virtually all non-Mendelian complex diseases are triggered by the additive effect of polygenicity and environmental factors. As the past decade has seen an explosion in GWA studies identifying genetic variants associated with disease risk, the next great challenge for researchers is to put a functional description on the mechanism driving these associations. This dissertation describes several imaging genetics studies with the aim of understanding how genetic risk variants influence multiple sclerosis and coronary artery disease, two hallmark complex polygenic disorders. Additionally, we describe several methods for quantifying “genetic connectivity” across traits and measurements. ❧ Multiple sclerosis is a neurodegenerative demyelinating disease with an unknown pathogenesis. The disease is highly heterogeneous with a highly complex genetic architecture. Examining the overlap between genetic loci that affect MS risk and regional brain structure may reveal mechanisms that influence lesion development in the brain. Based on recent genomic screens for common variants associated with MS, we examined pleiotropic overlap between summary statistics from the most recent MS case-control ImmunoChip study and GWAS summary statistics for seven subcortical brain volumes (nucleus accumbens, amygdala, caudate nucleus, globus pallidus, hippocampus, putamen, thalamus), and intracranial volume. We used continuous inflation analysis (CIA) to test profiles of overlap. We found significant evidence of overlap between variants in MS risk genes and the volumes of the hippocampus and thalamus. In follow up analyses of these structures, we used pleiotropy-informed conditional FDR and identified 24 novel variants associated with both MS risk and regional brain volumes. This discovery of gene variants associated with both MS risk and hippocampal and thalamic volumes in the brain may narrow the search for causal pathways mediating their effect on the human brain. ❧ The mechanism of how single nucleotide polymorphisms (SNPs) confer disease risk is not completely understood. MRI is utilized clinically and in research as a diagnostic tool in the identification of the hallmark white-matter lesions. We extracted gene expression maps of MS risk genes from the Allen Brain Atlas and overlaid them with results from Rinker et al., 2014 & 2015 where we showed an effect of several MS susceptibility alleles and polygenic risk score (PRS) on white-matter integrity in healthy young adults, and replicated these results in healthy aging adults. ❧ The degree to which shared genetic factors influence CAD and brain morphology will give important insight to the pathogenesis of CAD. Here, using SNP Effect Concordance Analysis (SECA) we combined summary statistics from the CARDIoGRAM GWAS with those from Hibar et al. 2015 GWAS of brain volumes, we tested for (1) global genetic pleiotropy (2) genetic concordance between genetic variants associated with CAD and brain volume. ❧ MRI and DTI measures of brain volume, cortical thickness and white matter (WM) integrity are commonly used in imaging genetics studies, but the genetic relationship between these measures is not well understood. Here we use structural equation models (SEM) in a twin design to test the genetic correlation between these common imaging measures. MRI and DTI data from 442 participants (mean age: 23.5 years +/- 2.1 SD

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

A multi-site neuroimaging approach to studying hippocampal damage in chronic stroke

PDF

Exploring the genetic basis of complex traits

PDF

Neuroimaging markers of risk & resilience to brain aging and dementia

PDF

Hispanic ethnicity in multiple sclerosis

PDF

Polygenic analyses of complex traits in complex populations

PDF

Identification of gene-exposure interactions for risk of cardiovascular disease

PDF

Unveiling the white matter microstructure in 22q11.2 deletion syndrome with diffusion magnetic resonance imaging

PDF

Pattern detection in medical imaging: pathology specific imaging contrast, features, and statistical models

PDF

Cross-sectional association of blood pressure, antihypertensive medications, MRI volumetric measures and cognitive function scores in an aging population

PDF

The contribution of arachidonate 5-lipoxygenase to inflammatory response in coronary artery disease

PDF

Environmental risk factors of Multiple Sclerosis: a twin study

PDF

Vascular contributions to brain aging along the Alzheimer's disease continuum

PDF

The association of cerebrovascular disease risk factors with brain structure and its modification by genetic variation

PDF

Examining the neuroanatomical correlates of reading in developmental dyslexia

PDF

Relationships between lifetime chronic stress exposure, vascular risk, cognition, and brain structure in HIV

PDF

Shortcomings of the genetic risk score in the analysis of disease-related quantitative traits

PDF

Understanding acute lymphoblastic leukemia in different ethnic groups in the United States

PDF

Representation problems in brain imaging

PDF

Independent and interactive effects of depression genetic risk and household socioeconomic status on emotional behavior and brain development

PDF

Identifying genetic, environmental, and lifestyle determinants of ethnic variation in risk of pancreatic cancer