Genetic contribution to multiple sclerosis risk among Ashkenazi Jews
Khankhanian et al. BMC Medical Genetics
Genetic contribution to multiple sclerosis risk among Ashkenazi Jews
Pouya Khankhanian 0
Takuya Matsushita 0
Lohith Madireddy 0
Antoine Lizée 0
Lennox Din 0
Jayaji M Moré 0
Pierre-Antoine Gourraud 0
Stephen L Hauser 0
Sergio E Baranzini 0
Jorge R Oksenberg 0
0 Department of Neurology, University of California, San Francisco , 675 Nelson Rising Lane, San Francisco, CA 94158 , USA
Background: Multiple sclerosis (MS) is an autoimmune disease of the central nervous system, with a strong genetic component. Over 100 genetic loci have been implicated in susceptibility to MS in European populations, the most prominent being the 15:01 allele of the HLA-DRB1 gene. The prevalence of MS is high in European populations including those of Ashkenazi origin, and low in African and Asian populations including those of Jewish origin. Methods: Here we identified and extracted a total of 213 Ashkenazi MS cases and 546 ethnically matched healthy control individuals from two previous genome-wide case-control association analyses, and 72 trios (affected proband and two unaffected parents) from a previous genome-wide transmission disequilibrium association study, using genetic data to define Ashkenazi. We compared the pattern of genetic risk between Ashkenazi and non-Ashkenazi Europeans. We also sought to identify novel Ashkenazi-specific risk loci by performing association tests on the subset of Ashkenazi cases, controls, probands, and parents from each study. Results: The HLA-DRB1*15:01 allele and the non-HLA risk alleles were present at relatively low frequencies among Ashkenazi and explained a smaller fraction of the population-level risk when compared to non-Ashkenazi Europeans. Alternative HLA susceptibility alleles were identified in an Ashkenazi-only association study, including HLA-A*68:02 and one or both genes in the HLA-B*38:01-HLA-C*12:03 haplotype. The genome-wide screen in Ashkenazi did not reveal any loci associated with MS risk. Conclusion: These results suggest that genetic susceptibility to MS in Ashkenazi Jews has not been as well established as that of non-Ashkenazi Europeans. This implies value in studying large well-characterized Ashkenazi populations to accelerate gene discovery in complex genetic diseases.
Multiple sclerosis; Ashkenazi jews; Genome-wide association study; Population genetics
Multiple sclerosis (MS) is an autoimmune disease with
central nervous system pathology  and the most
common cause of non-traumatic neurological disability in
young adults, affecting approximately 2.5 million people
worldwide. The global burden of MS has increased over
the past century but retains the well-known influence of
gender, latitude, and ancestry on risk. This is reflected in
the relatively high incidence in some population groups
(particularly those of European origin) compared with
others (African and Asian groups) [2, 3]. Likewise, high
frequency rates are found in Scandinavia, Iceland, the
British Isles and North America (about 1–2 in 1,000),
but lower frequencies are observed in the Mediterranean
Basin (0.5 in 1000).
There are notable exceptions to the European
prevalence gradient, such as in Sardinia where the prevalence
of MS is among the highest in the world (1.4/1,000) .
Similarly, Ashkenazi Jews in Israel (and the Diaspora)
are at high risk for developing MS . The distinctive
population histories of Sardinians and Ashkenazi Jews
include founder effects, admixture, and bottlenecks,
suggesting that unique genetic signatures underlie their
differential susceptibility. Unraveling these profiles may
provide important insights into the genetics of MS and
interactions with non-genetic factors.
Extensive empirical evidence confirms that genetic
variation is an important determinant of MS risk. Multiple
genome-wide association studies (GWAS) have been
completed and reported, including a multi-center effort with
nearly ten thousand cases . The classic HLA-DRB1 risk
locus within the MHC, specifically the HLA-DRB1*15:01
allele stood out in all GWAS with remarkably strong
statistical significance. In addition, 110 non-MHC variants
were found to be associated with disease susceptibility
(Additional file 1: Table S1) . As expected, each
identified variant conferred only modest odds ratios. These
studies have focused on datasets ascertained in Europe,
Australia and North America and included affected
Ashkenazi individuals. Jewish populations in the Diaspora
can be grouped into distinct genetic clades shaped by
admixture with local populations and social and cultural
forces. Nevertheless, they all maintain the ancestral
Eastern Mediterranean and Middle Eastern genomic
identity, which contrasts significantly with central and
northern European populations [8–12]. Here we seek
to clarify the genetic characteristics of MS in Ashkenazi
Previous studies reported the HLA-DRB1*15:01
association with MS in Ashkenazi Jews living in Israel, albeit
with reduced odds ratios compared to Europeans .
The HLA-DRB1*13:03 allele was the strongest genetic
biomarker of risk in a study of non-Ashkenazi Israelis 
but was not confirmed in Ashkenazi Israelis . In our
study, we use genomic signatures to identify and extract
Ashkenazi Jewish individuals from three SNP-chip-based
genome wide studies: a transmission disequilibrium (TD)
study and two case control studies. We next illustrate the
genetic relationship of the Ashkenazi subsets to the other
European sub-populations. We then compare allelic risk
at HLA-DRB1 and polygenic risk across the genome
between Ashkenazi Jews and non-Jewish western Europeans,
based on previously suggested risk alleles. To identify MS
susceptibility alleles that may be specific to the Ashkenazi
population, we perform case–control and TD tests for
SNPs across the genome and at classical HLA alleles in
the Ashkenazi subset.
We started by defining Ashkenazi subjects from three
previous multiple sclerosis studies: two genome-wide case–
control association studies (GENEMSA and WTCCC2)
and one genome-wide transmission disequilibrium (TD)
trio study (IMSGC). We describe the position of Ashkenazi
Europeans within the genetic landscape composed of other
European populations available in the WTCCC2 dataset.
We compare selected clinical characteristics of Ashkenazi
to non-Ashkenazi Europeans using available data in the
GENEMSA dataset. We then proceed to examine the MS
genetic risk burden in Ashkenazi using previously identified
MS susceptibility variants. Due to the unique properties of
the HLA risk compared to the genome-wide risk, we assess
the HLA risk before evaluating the genome-wide genetic
burden at 110 non-HLA MS susceptibility SNPs en masse
(using a 110-SNP MS genetic burden score, termed the
MSGB). We compare Ashkenazi cases to Ashkenazi
controls to determine whether previously identified risk alleles
apply to the Ashkenazi population. We compare MS risk
attributable to the previously identified genetic
susceptibility loci between Ashkenazi and non-Ashkenazi Europeans.
Finally, we seek to identify Ashkenazi-specific risk loci.
We perform case–control analyses in Ashkenazi subsets
of the GENEMSA and WTCCC2 studies and perform a
TD test of the Ashkenazi trios of the IMSGC study.
Defining the Ashkenazi subsets of previous studies
To extract individuals from case–control studies
(GENEMSA and WTCCC2) who had genetic similarity to a
well-defined Ashkenazi genetic reference dataset, we
built a hierarchy of clusters based on genome-wide IBD
distances. In the GENEMSA study, hierarchical clustering
revealed a large cluster of northern and western European
individuals (EUNW), a small cluster of eastern European
Ashkenazi Jews, and a small cluster of southern Europeans
(EUS). In the WTCCC2 study, there was a large cluster of
EUNW, a small cluster of European Ashkenazi Jews, and
a small cluster of Finnish samples. Based on the
hierarchical clustering, 97 Ashkenazi Jews and 1,541 EUNW in
GENEMSA and 662 Ashkenazi Jews and 26,487 EUNW
in WTCCC2 were identified (Fig. 1a). Multidimensional
scaling of genome-wide IBD distances (Fig. 1b) revealed a
large main cluster of EUNW and a smaller outlying
cluster of the well-characterized Ashkenazi controls in both
datasets. Structure-like analysis using FRAPPE  (a
maximum likelihood method to infer the genetic ancestry
of each individual, where the individuals are assumed to
have originated from K ancestral clusters) with K = 3 also
demarcated the cluster of Ashkenazi individuals (Fig. 1a).
To extract individuals of Ashkenazi origin from the
IMSGC trio study, multi-dimensional scaling was
performed as above. In this case, the first principal
component (PC1) was used to define Ashkenazi individuals
(Additional file 5). For each trio, the value of PC1 for the
proband was approximately the mean of the value of PC1
for the parents. Higher values of PC1 were defined as
Ashkenazi Jews and lower values were defined as EUNW.
For the purpose of the transmission disequilibrium test
(see meta-analysis section below), 76 Ashkenazi trios were
identified, where an Ashkenazi trio is defined as a trio
with at least one Ashkenazi parent. Approximately one
third (24 of 76) families had one non-Ashkenazi parent,
and in those families transmission from either parent was
considered for the analysis.
Fig 1 Hierarchical clustering and multi-dimensional scaling of individuals. a: Hierarchical clustering using genome-wide IBD in GENEMSA and the
WTCCC2 dataset. The dendrogram demonstrates relationship between individuals and the lower bars designate assigned population of each
individual. Control Ashkenazi Jews (AJ) corresponds to ethnically well-characterized Ashkenazi controls . The lower panel shows individual
ancestry and admixture proportions with K = 3. b: Multidimensional scaling of genome-wide IBD in GENEMSA and WTCCC2 dataset (pink =
previously well defined AJ controls , green = newly defined AJ cases and controls, yellow = EUNW, blue = EUS)
Detailed MS clinical characteristics of the Ashkenazi
with respect to Europeans were available for a small
proportion of the samples and are reported in the Additional
file 2. Some intriguing differences such as significantly
lower multiple sclerosis severity score in Ashkenazi, and
lower rates of motor weakness and acute transverse
myelitis require confirmation with larger datasets.
Relationship of Ashkenazi to other Europeans
The WTCCC2 MS dataset was collected in 15 different
countries spanning Europe, the US, and Australia. To
explore the relationship between Ashkenazi and other
populations of European ancestry, hierarchical clustering
of data from each country using whole-genome SNP
frequencies was performed (Fig. 2). This analysis showed
that Ashkenazi Jews grouped with Italian and Spanish
populations. The majority of the Ashkenazi study
participants came from UK and US populations. Populations
from Australia, Belgium, Denmark, France, Germany,
Sweden, the UK and the US (excluding the Ashkenazi
Jews) formed a larger cluster. The Finnish population was
relatively distant from the other European populations.
Two measures of linkage disequilibrium (D-prime and
r-squared) were plotted as a function of genetic distance
in base pairs for five groups of samples (Fig. 3). The
CEU in green is a reference set of northwestern
Europeans and the LWK and YRI samples in blue are two
reference sets of African samples, all from the 1000
genomes project (www.1000genomes.org). The
AJWTCCC2 and AJ-GENEMSA in red are two sets of
Ashkenazi samples from the current study. The same
number of individuals were used for all five groups. We
observed that the northwestern European population
exhibits greater LD than Ashkenazi and African
Fig 2 Hierarchical clustering of populations. A hierarchical clustering of cohorts that comprise the WTCCC2 dataset. The y-axis represents
genome-wide distance between populations (measured from genome-wide allele frequencies). Ashkenazi Jews cluster with the Spanish cases,
the Italian cases, and the Italian controls, as seen on the right. On the left, a large cluster of western European cases and controls from France,
Belgium, Germany, Australia, United States, and the United Kingdom. In the center, a cluster of northern European countries includes Sweden,
Denmark, Norway, and Germany. Two Finnish populations cluster tightly together. Three groups of relatively small sample size cluster loosely
including groups from Poland, Norway, and Ireland. For countries with two sets of controls (UK, Germany, Sweden), the cohort names are
provided in parentheses 
samples, suggesting that well powered Ashkenazi
datasets could add to fine mapping of regions of interests
identified in GWAS.
MS genetic burden of Ashkenazi: HLA
To search for HLA alleles associated with MS risk in
Ashkenazi, a case–control association study of imputed
HLA alleles was performed on the GENEMSA and
WTCCC2 datasets, and a genome-wide transmission
disequilibrium test was performed on the IMSGC
dataset. The HLA-DRB1*15:01 allele conferred risk in each
of the three studies, statistically significant in two of
three (Table 1). The HLA-DQB1*06:02 allele, highly
linked to HLA-DRB1*15:01, showed a similar pattern of
Fig 3 Linkage Disequilibrium. The median D-prime (y-axis, left panel), and median r-squared (y-axis, right panel), as a function of genetic distance
in base pairs (x-axis), for six groups of samples. CEU = reference northwestern European samples from Utah. LWK, YRI = reference sets of African
samples. AJ-WTCCC2, AJ-GENEMSA = two sets of Ashkenazi samples this study
aall Ashkenazi carriers of DQB1*06:02 were cases
risk. The HLA-DRB1*13:03 allele, previously implicated in
an a non-Ashkenazi Israeli MS dataset , was found in
only 11 total Ashkenazi individuals from the three datasets
and showed no evidence of association with disease. The
population risk attributable to a variant (Nagelkerke r2)
depends on the effect size (the odds ratio) and the frequency
of the risk variant in cases. While the odds ratio for the
HLA-DRB1*15:01 allele in Ashkenazi Jews is similar to that
in EUNW (Additional file 4), the HLA-DRB1*15:01 carrier
frequency is lower in Ashkenazi (for example, 15.6 % in
WTCCC Ashkenazi controls versus 28.1 % in WTCCC
EUNW controls) (Additional file 4). It is therefore not
surprising that the risk explained by the HLA-DRB1*15:01
allele is lower in Ashkenazi Jews (GENEMSA Nagelkerke r2
= 0.009, WTCCC2 r2 < 0.0001) than in EUNW (GENEMSA
r2 = 0.145, WTCCC2 r2 = 0.11) (Fig. 4).
Three other HLA alleles showed nominal level of
association (p < 0.05, uncorrected for multiple hypothesis
testing) in a meta-analysis of the three datasets. The
HLAA*68:02 showed positive association with MS (p = 0.04),
while the HLA-B*38:01 and HLA-C*12:03 alleles showed
negative association (p = 0.01 and 0.0001 respectively).
A step-wise conditional analysis (Fig. 5) was performed
to evaluate statistically significant independent HLA
effects. In the first analysis (Fig. 5 panel a), SNPs from
across MHC class I and class II (panel A, left side) passed
the Bonferroni threshold of 0.05 (red line on figure). The
first significant HLA allele (most significant p-value),
came from the HLA-C gene (panel A right side) and was
the C*12:03 allele (thought the allele does not pass
Bonferroni correction). In the second analysis,
HLAC*12:03 was included as a covariate and we looked for an
independent effect (panel B). Here again a handful of
SNPs passed the Bonferroni threshold, and the most
significant finding was at the HLA-A gene, where the
A*68:02 allele was most significant. In the third analysis,
HLA-C*12:03 and HLA-A*68:02 were both included as
covariates (panel C). At this stage, only 3 SNPs passed the
Bonferroni cutoff, and the top finding was the
HLADRB1*15:01 allele. In the fourth and final analysis, the
covariates were HLA-C*12:03, HLA-A*68:02, and
HLADRB1*15:01 (Panel D); no further statistically significant
SNP or HLA allele associations were seen. Of note, in
each step of the analysis, the top HLA allele was chosen
as a covariate for the next step, even if a single SNP may
have had higher significance than the HLA allele, and even
if the HLA allele did not pass Bonferroni threshold.
MS genetic burden of Ashkenazi: Genome-wide non-HLA
110 genome-wide SNPs and a polygenic risk score
(MSGB) were used to assess genetic risk of MS. The
mean MSGB was significantly higher in Ashkenazi cases
than Ashkenazi controls (WTCCC2: difference d = 0.31,
p = 3.5*10−7; GENEMSA d = 0.30, p = 0.017) (Fig. 6). As
expected, a similar difference (d) was seen between
EUNW cases and EUNW controls (WTCCC2: d = 0.37,
p < 2.2*10−16; GENEMSA: d = 0.32, p < 2.2*10−16) (Fig. 6).
Following the pattern observed for HLA risk, the
nonHLA MSGB was lower in Ashkenazi cases than EUNW
cases in both datasets (WTCCC2: d = 0.14, p = 0.0029;
GENEMSA: d = 0.05, p = 0.7) but was statistically
significant in only one of two datasets. The non-HLA MSGB
was also lower in Ashkenazi controls than in the EUNW
controls (WTCCC2: d = 0.08, p = 0.0022; GENEMSA: d
= 0.03, p = 0.7) but was statistically significant in only
one of the two datasets. Therefore, it is again not
surprising that the percent of risk explained by the MSGB
is lower in Ashkenazi (GENEMSA r2 = 0.024, WTCCC2
r2 = 0.011) than in non-Ashkenazi Europeans
(GENEMSA r2 = 0.068, WTCCC2 r2 = 0.087).
Earlier, we have shown using a genome-wide
identityby-descent metric that within the country-based
WTCCC2 dataset, the nearest European populations to
the Ashkenazi Jews are Spanish and Italian cases and
controls (Fig. 2), but the Ashkenazi cases had
significantly lower non-HLA MSGB compared to the Italian
and Spanish (cases (p = 0.002), whereas Italian and
Spanish cases showed no significant difference from the
EUNW cases. The Ashkenazi Jews controls had lower
non-HLA MSGB than controls from Italy (p = 0.04).
Controls from Italy showed no significant difference
from the EUNW controls. These observations, altogether,
suggest a larger missing heritability in Ashkenazi MS. To
screen the genome for genetic susceptibility to MS in an
all-Ashkenazi dataset, a genome-wide case–control
association study of SNPs was performed on the GENEMSA
(n = 53 cases, 42 controls) and WTCCC2 (n = 136 cases,
429 controls) datasets, and a genome-wide transmission
disequilibrium test was performed on the IMSGC dataset
(n = 76 trios). There were no significant results at a
genome-wide FDR < 0.1 in any study and the
metaanalysis of the three studies yielded no significant results
at a genome-wide FDR < 0.1, which is not surprising
given the limited power of this dataset.
Fig 4 Multiple sclerosis risk attributed to genetics. The Multiple sclerosis risk (Nagelkerke R2) attributable to HLA-DRB1*15:01 and the non-HLA
MSGB score in the GENEMSA and WTCCC2 datasets
A 2006 survey of MS frequency in Israel reported that the
highest disease rates were in Israeli-born Jews and in Jewish
immigrants from Europe/America, with prevalence similar
to that seen in Europeans . Jewish immigrants from
African/Asian countries and Christian Arabs had
intermediate MS rates (significantly lower than in the first two
groups but not significantly different from each other) .
Moslem Arabs, Druze, and Bedouins had the lowest rates
of MS. Karni and colleagues  highlight an intriguing
difference in prevalence between Jewish immigrants from
Africa/Asian countries (low prevalence) and their Israeli
born children (higher prevalence) suggesting a strong
environmental influence acting across a single generation
[15, 16]. The HLA-DRB1*15:01 allele has been shown to
be a risk allele in Ashkenazi , but it is less frequent in
Ashkenazi (~5 %) [11, 16] compared to the general
European population (~15-20 %), which is a
counterintuitive observation given the high frequency of MS among
In this study, we have extracted Ashkenazi individuals
using genetic data from three datasets. A key limitation of
this study is the sample size, the small number of
identified Ashkenazi individuals leaves little power for discovery
of new variants, especially at the genome-wide level.
Another limitation the small sample size of the
wellcharacterized Ashkenazi controls  used to help define
Ashkenazi from within our MS datasets. A larger set of
well-characterized controls may allow the identification of
more Ashkenazi individuals, and may make the
identification more reproducible across future studies.
Despite the power limitation, we have confirmed that
the classical MS determinant HLA-DRB1*15:01 is a risk
allele in Ashkenazi. SNP-based HLA analysis using
validated imputed techniques revealed the well-known
HLADRB1*15:01- HLA-DQB1*06:02 association with risk was
significant in Ashkenazi, but its frequency in Ashkenazi
cases is significantly lower than the frequency in other
European cases. The linked class I HLA-B*38:01 and
HLA-C*12:03 alleles showed a nominally significant
protective effect, which is in line with a previously described
protective effect of HLA-B*38:01 in a very large European
study . Also relevant, a HLA-B38 protective effect was
detected in an Iranian MS population . The
HLAA*68:02 allele showed a nominally significant risk effect.
This allele belongs to multiple haplotypes  and is
linked to previously described class II risk alleles,
HLADRB1*13:01 [6, 13], HLA-DRB1*03:01  and
HLADRB1*15:03 in African Americans .
To our knowledge, this is the first study to validate HLA
imputation performed on Ashkenazi samples using a
European reference population to train the model. We
have seen that generally imputation works well for most
alleles of most HLA genes, but works very poorly for
specific alleles (for example, DRB1*11:01, see Additional file 3
for further examples), which represents a limitation of this
study. Other studies using HLA imputation to derive
HLA alleles for Ashkenazi would likely benefit from
Fig 5 Human Leukocyte Antigen Alleles associated with multiple sclerosis in Ashkenazi Jews. A step-wise conditional analysis was performed to
evaluate statistically significant independent HLA effects. Panel a left: SNPs from across MHC class I and class II passed the Bonferroni threshold of
0.05 (red line on figure), the top hit coming from class II near DRB1 and DQB1. Panel a right: The top significant HLA allele came from the HLA-C
gene and was the *12:03 allele. Panel b: In the second analysis, HLA-C*12:03 was included as a covariate. Here again a handful of SNPs passed the
Bonferroni threshold, and the top significant finding was at the HLA-A gene, where the *68:02 allele was most significant. Panel c: In the third
analysis, HLA-C*12:03 and HLA-A*68:02 were both included as covariates. Here 3 SNPs passed the Bonferroni cutoff, and the top finding was the
*15:01 allele at HLA-DRB1. Panel d: In the fourth analysis, the covariates were HLA-C*12:03, HLA-A*68:02, and HLA-DRB1*15:01 and no further
statistically significant associations were seen
masking those specific alleles, rather than trying to set a
single threshold on the confidence metric of imputation.
Ideally, an all-Ashkenazi reference population would be
used to train the model.
The historical record shows that Jewish people
emigrated in mass from their ancestral home in the Eastern
Mediterranean, beginning over two and a half millennia
ago, establishing Jewish communities in many different
regions across the globe. The contemporaneous Jewish
population is generally grouped according to the most
recent place of origin into two main groups, Ashkenazi
(originating from Eastern, Central, and Northern Europe),
and Non-Ashkenazi (from North Africa, the Middle East
and Asia). Members of each group differ in physiognomy
and life style suggesting significant admixture as the force
driving Jewish population diversity. However, recent
genome-wide assessment of multiple Jewish datasets note
their considerable degree of genetic homogeneity and
closeness to other populations of the Levant, especially the
Druze and Palestinians. Our own data using primarily
UK and American Ashkenazi genomes suggests a
degree of genome-wide similarity between Ashkenazi
and Mediterranean in the context of a primarily
northwest European cohort consistent with previous
findings. Portions of genetic susceptibility to MS in
Ashkenazi are shared with Europeans (HLA-DRB1*15:01
for example), while others are shared with another middle
eastern population (HLA-B*38 and HLA-C*12 for
example), consistent with their previously reported shared
ancestry [9, 12].
'The non-HLA polygenic risk score conferring risk in
Ashkenazi was lower in Ashkenazi cases than European
cases, and, altogether, the previously described and
validated risk alleles (DRB1*15:01 and genome-wide) explain
a relatively smaller fraction of the genetic susceptibility to
MS in Ashkenazi. Comparable differences were found
between Ashkenazi controls and European controls,
indicating that underlying differences in the healthy population
can explain the apparent genomic differences in cases.
The precise reason for the decreased concentration of MS
susceptibility alleles in Ashkenazi Jews is unknown, and
additional research is necessary to resolve the effects of
Fig 6 Multiple sclerosis genetic burden. Multiple sclerosis genetic burden in multiple sclerosis patients and controls of Ashkenazi Jewish (AJ)
origin and Europeans (EUNW). The mean MSGB for each group is displayed above the graph. The number of samples in each group is displayed
beside each box-plot
selection and drift in the context of cultural isolation,
admixture and migration. In this study, we also attempted to
perform disease association analyses in a small Ashkenazi
cohort. We found no evidence of SNP-level association
using a genome-wide SNP-based analysis due most likely to
lack of power, a rather important limitation of this study.
Well-powered GWAS, re-sequencing and epidemiological
studies in Ashkenazi datasets may provide a unique
opportunity to further decipher the genetic and
geneenvironment underpinnings of MS.
Sources of data
The WTCCC2 dataset was provided by a collaboration
between the International MS Genetics Consortium and the
Wellcome Trust Case Control Consortium. This GWAS
dataset consisted of 30,248 individuals . This included
11,376 cases diagnosed with MS, as well as 18,872
controls, all of European ancestry. Cases were collected from
reference centers in Australia, Belgium, Denmark, Finland,
France, Germany, Ireland, Italy, New Zealand, Norway,
Poland, Spain, Sweden, the UK, and the USA. Controls
were collected from 12 centers. Genotyping was
performed using Illumina Human 660-Quad chip in cases
and Illumina Human 1.2 M - Duo chip in controls.
Extensive quality control analyses were performed to reduce
population stratification and inflation.
The GENEMSA dataset included 1,857 European
ancestry individuals from a case–control GWAS  and
consisted of 975 cases and 882, age- and gender-matched
controls. Cases and controls were collected from reference
centers in the USA, Netherlands, and Switzerland.
Genotyping was performed with Illumina Human Hap 550 K
The IMSGC dataset was provided by the International
MS Genetics Consortium. This trio dataset included
2,790 European ancestry individuals from a family-based
genome-wide study and consisted of 930 MS cases and
their unaffected parents . Trios were collected at
reference sites in the USA and the UK. Genotyping was
performed using Affymetrix GeneChip Human Mapping
500 K arrays.
In all datasets, MS cases were diagnosed by
neurologists specialized in demyelinating diseases in accordance
with well-established diagnostic and study inclusion
criteria . Study protocols were approved by the local
Committees on Human Research and Informed consent
was obtained from all participants.
Defining European Jewish ancestry
Europeans can be stratified into ancestral subgroups using
a set of SNP markers from across the genome [8, 24–27].
In order to identify Ashkenazi individuals included within
the three study datasets, genome-wide SNP data from a
set of ethnically well-characterized Ashkenazi controls 
were merged with the study datasets. After removing
SNPs with minor allele frequency < 1 % and SNPs with >
0.1 % missing calls, 47,251 independent SNPs (with
pairwise r2 < 0.1) from the merged dataset were used to
calculated identity by descent (IBD) using software Beagle ,
aggregated over ten rounds. Multidimensional scaling 
(MDS) with 50 dimensions and hierarchical Ward 
clustering using IBD distances was performed using R
software. Definition of Ashkenazi was performed
separately in each of the three datasets (GENEMSA, WTCCC2,
IMSGC). Ashkenazi ancestry was defined using
hierarchical clustering in the case–control datasets (GENEMSA
and WTCCC2) and using MDS in the trio study (IMSGC)
due to the inherent interrelatedness of individuals in the
In each of the three datasets, SNP information was used
to impute classical HLA alleles at 5 loci: A, B, C, DRB1,
and DQB1. The open-source R package HiBAG software
was used for imputation. Briefly, an ensemble classifier
is created consisting of individual classifiers with HLA
and SNP haplotype probabilities estimated from
bootstrapped samples and SNP subsets and HLA type
predictions are averaged over the posterior probabilities
from all classifiers. The published model for populations
of European ancestry was used. Validation of HLA
imputation was performed on a subset of 806 Europeans
and 94 Ashkenazi from the GENEMSA dataset using
sequence-based genotyping. Due to differences in LD
decay and allele frequency between Ashkenazi and
Europeans, quality control was performed at each allele of
each locus in order to minimize the error rate while
maximizing the call rate (see Additional file 3).
Polygenic risk score
As of July 2014, SNPs at 110 non-HLA variants had
confirmed associations with MS [7, 31] (Additional file 1:
Table S1), each with a much weaker conferred risk (odds
ratio ≈ 1.2-1.4) than HLA-DRB1*1501. For each individual,
a multiple sclerosis genetic burden, MSGB , was
calculated across these non-HLA genes. Briefly, the MSGB
is a polygenic risk score, the weighted sum of the number
of risk alleles carried, where weights are determined by log
of published odds ratios. Differences of MSGB score
between cases and controls were tested by one-sided
Wilcoxon test because the MSGB score consists of risk
alleles for MS.
Disease association studies in Ashkenazi Jews
To look for MS genetic susceptibility loci that are
specific to Ashkenazi Jews, Ashkenazi Jewish individuals
from each of the three datasets were extracted. A
metaanalysis genome-wide study on all SNPs and a
metaanalysis of imputed HLA alleles was then performed.
For SNPs in the WTCCC2 study and the GENEMSA
study, genome-wide logistic regression was performed
with age, gender, and PCA covariates. For HLA alleles in
the GENEMSA and WTCCC studies, genotypes of each
allele (e.g. HLA-DRB1*15:01) were re-coded as 0, 1, or 2
copies and logistic regression was performed. For SNPs
in the IMSGC trio study, a transmission disequilibrium
(TD) test was performed. For HLA alleles in the IMSGC
study, genotypes were re-coded as 0, 1, or 2 copies and
at TD test was performed. Logistic regression, TD tests,
and random-effects meta-analysis were implemented in
Additional file 1: Table S1. The 110 previously identified variants
associated with MS, and the nearest genes. (DOC 166 kb)
Additional file 2: Clinical Characteristics of Ashkenazi. (DOCX 39 kb)
Additional file 3: HLA imputation. (DOC 819 kb)
Additional file 4: Figure S1. Principal component analysis of the IMSGC
trio dataset using genome-wide IBD distances, showing component 1
(x-axis) versus component 2 (y-axis). The first principal component (PC1)
differentiates Ashkenazi (higher values) from other Europeans (lower
values). Probands are highlighted as shaded red circles (females) or
squared (males). Unaffected parents are highlighted as open red circles
(mothers) and squares (fathers). Parents are connected to probands by
dashed lines. In each successive picture a different set of trios is
highlighted (since highlighting all trios at once makes the figure
unreadable). Panels 2 thru 6 demonstrate probands of partial Jewish
Ancestry (intermediate values of the PC-1 axis) may have a single Jewish
parent and a single non-Jewish parent, or they may have two parents
each of partial Jewish ancestry. (XLSX 191 kb)
Additional file 5: Supplementary figure PCA of IMSGC. (PDF 230 kb)
Genetic data were provided by the International Multiple Sclerosis Genetics
Consortium, the GeneMSA Consortium, and the Wellcome Trust
Case–control Consortium. This study was supported by the National
Institutes of Health (RO1NS26799 and R01NS49477). The funding body did
not participate in collection, analysis, and interpretation of data, in the
writing of the manuscript, or in the decision to submit the manuscript for
1. Hauser SL , Goodin DS : Multiple Sclerosis and Other Demyelinating Diseases, 17th edition edn: McGraw Hill ; 2010 .
2. Alonso A , Hernan MA . Temporal trends in the incidence of multiple sclerosis: a systematic review . Neurology . 2008 ; 71 ( 2 ): 129 - 35 .
3. Sotgiu S , Pugliatti M , Sotgiu A , Sanna A , Rosati G . Does the “hygiene hypothesis” provide an explanation for the high prevalence of multiple sclerosis in Sardinia? Autoimmunity . 2003 ; 36 ( 5 ): 257 - 60 .
4. Granieri E , Casetta I , Govoni V , Tola MR , Marchi D , Murgia SB , et al. The increasing incidence and prevalence of MS in a Sardinian province . Neurology . 2000 ; 55 ( 6 ): 842 - 8 .
5. Alter M , Kahana E , Zilber N , Miller A. Multiple sclerosis frequency in Israel's diverse populations . Neurology . 2006 ; 66 ( 7 ): 1061 - 6 .
6. International Multiple Sclerosis Genetics Consortium (IMSGC), Wellcome Trust Case Control Consortium 2 (WTCCC2), Sawcer S , Hellenthal G , Pirinen M , Spencer CC , et al. Genetic risk and a primary role for cell-mediated immune mechanisms in multiple sclerosis . Nature . 2011 ; 476 ( 7359 ): 214 - 9 .
7. International Multiple Sclerosis Genetics Consortium (IMSGC) , Beecham AH , Patsopoulos NA , Xifara DK , Davis MF , Kemppinen A , et al. Analysis of immune-related loci identifies 48 new susceptibility variants for multiple sclerosis . Nat Genet . 2013 ; 45 ( 11 ): 1353 - 60 .
8. Atzmon G , Hao L , Pe'er I , Velez C , Pearlman A , Palamara PF , et al. Abraham's children in the genome era: major Jewish diaspora populations comprise distinct genetic clusters with shared Middle Eastern Ancestry . Am J Hum Genet . 2011 ; 86 ( 6 ): 850 - 9 .
9. Behar DM , Yunusbayev B , Metspalu M , Metspalu E , Rosset S , Parik J , et al. The genome-wide structure of the Jewish people . Nature . 2010 ; 466 ( 7303 ): 238 - 42 .
10. Hammer MF , Redd AJ , Wood ET , Bonner MR , Jarjanazi H , Karafet T , et al. Jewish and Middle Eastern non-Jewish populations share a common pool of Y-chromosome biallelic haplotypes . Proc Natl Acad Sci U S A . 2000 ; 97 ( 12 ): 6769 - 74 .
11. Klitz W , Gragert L , Maiers M , Fernandez-Vina M , Ben-Naeh Y , Benedek G , et al. Genetic differentiation of Jewish populations . Tissue Antigens . 2010 ; 76 ( 6 ): 442 - 58 .
12. Carmi S , Hui KY , Kochav E , Liu X , Xue J , Grady F , et al. Sequencing an Ashkenazi reference panel supports population-targeted personal genomics and illuminates Jewish and European origins . Nat Commun . 2014 ; 5 : 4835 .
13. Kwon OJ , Karni A , Israel S , Brautbar C , Amar A , Meiner Z , et al. HLA class II susceptibility to multiple sclerosis among Ashkenazi and non-Ashkenazi Jews . Arch Neurol . 1999 ; 56 ( 5 ): 555 - 60 .
14. Tang H , Peng J , Wang P , Risch NJ . Estimation of individual admixture: analytical and study design considerations . Genet Epidemiol . 2005 ; 28 ( 4 ): 289 - 301 .
15. Karni A , Kahana E , Zilber N , Abramsky O , Alter M , Karussis D. The frequency of multiple sclerosis in Jewish and Arab populations in greater Jerusalem . Neuroepidemiology. 2003 ; 22 ( 1 ): 82 - 6 .
16. Karni A , Kohn Y , Safirman C , Abramsky O , Barcellos L , Oksenberg JR , et al. Evidence for the genetic role of human leukocyte antigens in low frequency DRB1*1501 multiple sclerosis patients in Israel . Mult Scler . 1999 ; 5 ( 6 ): 410 - 5 .
17. Patsopoulos NA , Barcellos LF , Hintzen RQ , Schaefer C , van Duijn CM , Noble JA , et al. Fine-mapping the genetic association of the major histocompatibility complex in multiple sclerosis: HLA and non-HLA effects . PLoS Genet . 2013 ; 9 ( 11 ): e1003926 .
18. Amirzargar A , Mytilineos J , Yousefipour A , Farjadian S , Scherer S , Opelz G , et al. HLA class II (DRB1, DQA1 and DQB1) associated genetic susceptibility in Iranian multiple sclerosis (MS) patients . Eur J Immunogenet . 1998 ; 25 ( 4 ): 297 - 301 .
19. Gourraud PA , Harbo HF , Hauser SL , Baranzini SE . The genetics of multiple sclerosis: an up-to-date review . Immunol Rev . 2012 ; 248 ( 1 ): 87 - 103 .
20. Oksenberg JR , Barcellos LF , Cree BA , Baranzini SE , Bugawan TL , Khan O , et al. Mapping multiple sclerosis susceptibility to the HLA-DR locus in African Americans . Am J Hum Genet . 2004 ; 74 ( 1 ): 160 - 7 .
21. Baranzini SE , Wang J , Gibson RA , Galwey N , Naegelin Y , Barkhof F , et al. Genome-wide association analysis of susceptibility and clinical phenotype in multiple sclerosis . Hum Mol Genet . 2009 ; 18 ( 4 ): 767 - 78 .
22. Hafler DA , Compston A , Sawcer S , Lander ES , Daly MJ , De Jager PL , et al. Risk alleles for multiple sclerosis identified by a genomewide study . N Engl J Med . 2007 ; 357 ( 9 ): 851 - 62 .
23. Polman CH , Reingold SC , Edan G , Filippi M , Hartung HP , Kappos L , et al. Diagnostic criteria for multiple sclerosis: 2005 revisions to the “McDonald Criteria” . Ann Neurol . 2005 ; 58 ( 6 ): 840 - 6 .
24. Gusev A , Palamara PF , Aponte G , Zhuang Z , Darvasi A , Gregersen P , et al. The Architecture of Long-Range Haplotypes Shared within and across Populations . Mol Biol Evol . 2012 ; 29 ( 2 ): 473 - 86 .
25. Need AC , Kasperaviciute D , Cirulli ET , Goldstein DB. A genome-wide genetic signature of Jewish ancestry perfectly separates individuals with and without full Jewish ancestry in a large random sample of European Americans . Genome Biol . 2009 ; 10 ( 1 ): R7 .
26. Nelson MR , Bryc K , King KS , Indap A , Boyko AR , Novembre J , et al. The Population Reference Sample, POPRES: a resource for population, disease, and pharmacological genetics research . Am J Hum Genet . 2008 ; 83 ( 3 ): 347 - 58 .
27. Price AL , Butler J , Patterson N , Capelli C , Pascali VL , Scarnicci F , et al. Discerning the ancestry of European Americans in genetic association studies . PLoS Genet . 2008 ; 4 ( 1 ): e236 .
28. Browning BL , Browning SR. A fast, powerful method for detecting identity by descent . Am J Hum Genet . 2011 ; 88 ( 2 ): 173 - 82 .
29. Gower JC . Some distance properties of latent root and vector methods used in multivariate analysis . Biometrika . 1966 ; 53 : 325 - 8 .
30. Hartigan JA . Clustering Algorithms . New York : Wiley; 1975 .
31. Gourraud PA , McElroy J , Caillier SJ , Johnson BA , Santaniello A , Hauser SL , et al. Aggregation of multiple sclerosis genetic risk variants in multiple and single case families . Ann Neurol . 2011 ; 69 ( 1 ): 65 - 74 .