Genes and Immunity (2010) 11, 423–431 & 2010 Macmillan Publishers Limited All rights reserved 1466-4879/10 www.nature.com/gene

ORIGINAL ARTICLE Opposing effects of the HLA-DRB1*0301-DQB1*0201 haplotype on the risk for multiple sclerosis in diverse Arab populations in

G Benedek1,6, T Paperna2,6, N Avidan2, I Lejbkowicz2, JR Oksenberg3, J Wang3, C Brautbar1,4, S Israel4 and A Miller2,5, for the Israeli MS Genetics group7 1The Lautenberg Center for Immunology, Hebrew University Hadassah Medical School, , Israel; 2Faculty of Medicine, Rappaport Research Institute, Technion–Israel Institute of Technology, , Israel; 3Department of Neurology, University of California, San Francisco, CA, USA; 4Tissue Typing and Immunogenetics Unit, , Jerusalem, Israel and 5Division of Neuroimmunology and Multiple Sclerosis Center, Carmel Medical Center, Haifa, Israel

Different multiple sclerosis (MS) prevalence rates were reported for Muslim and Christian Arabs in Israel. In this study, we evaluated whether associations of human leukocyte antigen (HLA) genes with MS may contribute to this prevalence difference. DNA samples from Israeli Arab MS patients (n ¼ 109) and controls (n ¼ 132) were typed for HLA class I (HLA-A, -B and -C) and II (HLA-DRB1 and -DQB1) genes. Global comparisons of HLA allele frequencies revealed significant differences between Christians and Muslims; therefore, case–control analyses were stratified by religious affiliation. Disease characteristics of Muslim and Christian Arab MS patients were similar to those reported for European populations. Opposing association signals with MS were observed for alleles composing the DRB1*0301-DQB1*0201 haplotype: positive association of the HLA-

DRB1*0301 allele in Muslims (PBonferroni ¼ 0.004, odds ratio (OR) ¼ 3.07), and negative association in Christian Arabs

(PBonferroni ¼ 0.01, OR ¼ 0.12), with similar results obtained for HLA-DQB1*0201. HLA-B*52 was negatively associated with MS only in Muslims (PBonferroni ¼ 0.01, OR ¼ 0.03). The study presents for the first time a high-resolution HLA gene analysis in clinically well-characterized Arab populations with MS, and shows the population-specific contribution of the DRB1*0301- DQB1*0201 haplotype to disease susceptibility. Genes and Immunity (2010) 11, 423–431; doi:10.1038/gene.2010.20; published online 13 May 2010

Keywords: association studies; autoimmune disease; human leukocyte antigen; major histocompatibility complex; multiple sclerosis; population genetics

Introduction Most genetic studies in MS focused, not surprisingly, on populations of Northern European descent, where Multiple sclerosis (MS) is a neurological disease with a prevalence of MS is relatively high. These studies have strong inflammatory component. The precise etiology repeatedly affirmed the contribution of the HLA DR2 of MS remains unknown; however, both genetic and haplotype DQB1*0602-DQA1*0102-DRB1*1501-DRB5*0101 environmental factors presumably underlie disease to disease risk; however, further dissection of the specific pathogenesis. The major genetic effect has been attrib- role of the genes within this haplotype has been limited uted to the 6p major histocompatibility complex region, by the high linkage disequilibrium structure prevalent in comprising the human leukocyte antigen (HLA) genes.1,2 this chromosomal region. Recognition of this limitation Within this region, HLA class II genes such as HLA- has provoked studies in non-European populations, DRB1 and HLA-DQB1 appear to be important in taking advantage of their different fine genomic archi- susceptibility and progression of disease, but a growing tecture. For example, studies in African-American body of evidence is consistent with contributions beyond cohorts confirmed the contribution of DRB1*0301 and the HLA class II region, including effects from the nearby DRB1*1503 to MS risk, and the influence of HLA-DRB5 HLA class I genes HLA-A, -B and -C.3–6 on disease progression.7–9 In addition, studies on populations with unique MS prevalence patterns, or lower rates of the major susceptibility DR2 haplotype Correspondence: Dr A Miller, Division of Neuroimmunology and have revealed the full range of allelic heterogeneity at the Multiple Sclerosis Center, Carmel Medical Center, 7 Michal Street, DRB1 locus. Haifa 34362, Israel. Epidemiology studies suggest a relation between MS E-mails: [email protected] or [email protected] 6 prevalence and latitude, with risk increasing with distance These authors contributed equally to this work. 10 7See Appendix. from the equator. However, specific enclaves of high or Received 29 October 2009; revised 29 January 2010; accepted low risk have been reported in some populations, like 11 February 2010; published online 13 May 2010 Sardinia.11 The presence of population-specific factors, Genetics of multiple sclerosis in Israeli Arabs G Benedek et al 424 including founder effects, particularly in immune-related Table 1 Demographic data of participants and clinical character- genes, may modulate MS risk in different populations istics of MS patients residing in the same geographical region.12–15 Middle East Arab populations have a range of MS prevalence in the Muslims Christians low to medium rates.16–19 A recent epidemiological survey in Israel revealed that prevalence rates for Christian Arabs N (Cases/Controls) 66/90 43/42 are thrice as high as for Muslim Arabs, suggesting that Female/male (% females) Cases 46/20 (70) 29/14 (67) different risk factors operate in these two subethnic Controls 37/53 (41) 17/25 (40) groups.19 The Christian Arab population is a minority group among the Arab populations (B1.7% of the total Age at recruitment, years population in Israel versus 16.6% for Muslim Arabs, Mean±s.d. (range) according to the 2008 census, http://www1.cbs.gov.il/ Cases 37.6±10.7 (19–58) 38.3±11.6 (18–62) ± ± shnaton59/st02_19.pdf). It has therefore been the subject of Controls 37.7 11.0 (19–71) 40.0 14.3 (19–76) few genetic studies and little is known on its population Age of onset, years structure. It is not clear whether these MS prevalence Mean±s.d. (range) 30.7±9.2 (15–54) 29.7±9.6 (13–53) differences between Christians and other Israeli Arabs are the result of genetic differences, environmental effects or Disease duration from onset to recruitment, years both. Because religion in the Middle East Arab population Median (range) 5.5 (0–34) 6.0 (1–23) ± ± ± has shaped marriage practice over generations, it is likely Mean s.d. 6.8 5.8 8.6 5.9 that relatively high genetic homogeneity is present, which Years from onset to diagnosis may be advantageous for genetic studies as it is expected Median (range) 1.0 (0–10) 1.0 (0–20) to increase the statistical power.20 Mean±s.d. 2.1±2.7 3.4±5.0 Accordingly, two motives served as the driving force for this study: first, the advantage of genetic studies in EDSS at recruitment, N (%) non-European, inbred populations, where different p3 25 (37.9) 23 (53.5) 3o and o6 32 (48.5) 17 (39.5) linkage disequilibrium structure may lead to unmasking 6p 9 (13.6) 3 (7.0) of new genetics factors for MS; and second, to provide clinical and epidemiological data on specific Arab Disease type, N (%) populations for which there is relatively scarce informa- Relapsing-remitting 43 (65.2) 34 (79.1) tion. Identification of distinct genetic profiles associated Secondary-progressive 11 (16.7) 5 (11.6) with MS may explain the different prevalence rates for Primary-progressive 3 (4.5) 0 Relapsing-progressive 9 (13.6) 4 (9.3) MS in Muslims and Christian Arabs residing in Israel. Patients with relapsing 61 (96.8) 38 (88.4) onset ever treated with Results immunomodulatory treatment, N (%)a Demographic data of the study participants and clinical characteristics are presented in Table 1. The female/male Abbreviation: EDSS, expanded disability status scale. ratio and the age of onset in the two patient populations aInterferon-b-1a, interferon-b-1b or glatiramer acetate, as recorded (Muslim and Christian Arabs) were similar, as were the at the day of recruitment. other clinical characteristics, including disease duration, and access to disease-modifying treatments. Many of the individuals serving as controls were the spouses of the Table 2 P-values for allelic exact tests comparing Muslim and patients. This enrollment strategy was chosen to increase Christian cohorts for HLA genesa similarity between cases and controls regarding domi- ciles and lifestyle factors; therefore gender ratio was HLA-DQB1 HLA-DRB1 HLA-B HLA-C HLA-A significantly different between the case and control groups (Pp0.01). Case and control groups were well Cases 0.002 0.001 0.02 0.008 0.42 matched for age in both ethnic groups. Controls 0.31 0.53 0.30 0.79 0.03

HLA allele frequencies in the Israeli Muslim and Christian aGlobal w2 exact tests compared between religion groups in either Arab populations case groups or controls. Global P-values are shown; significant P Global comparisons of allele frequency distributions values (o0.05) shown in gray. revealed significant differences between Christians and Muslims among the cases for HLA-DQB1,-DRB1,-B and (Table 3; P ¼ 0.005 and 0.047, respectively). Further -C (P ¼ 0.002, 0.001, 0.02 and 0.008), and for HLA-A in the analysis indicated that in the Muslim population, control group (P ¼ 0.03) (Table 2); therefore, case–control DRB1*0301 is a major susceptibility allele, positively

analyses were stratified by religious affiliation. Tables 3, associated to MS (PB ¼ 0.004, odds ratio (OR) ¼ 3.07). 5–7 present the allele frequencies for the five HLA loci Although the previously established risk allele studied. Deviation from Hardy–Weinberg equilibrium DRB1*1501 was more frequent in the patient group, this was not observed for any of the HLA gene alleles tested. difference did not attain statistical significance. Surpris- ingly, in the Christian Arab population the DRB1*0301

HLA class II analyses allele was negatively associated with MS (PB ¼ 0.012, Case–control analysis was significant for the HLA-DRB1 OR ¼ 0.12). Moreover, DRB1*0301 homozygotes were gene in both the Muslim and the Christian populations observed only in the Christian control group (Table 4).

Genes and Immunity Genetics of multiple sclerosis in Israeli Arabs G Benedek et al 425 Table 3 HLA DRB1 percent allele frequencies

DRB1 Muslims Christians

a a Cases Controls PB OR Cases Controls PB OR (2N ¼ 132) (2N ¼ 180) (nominal) (95% CI) (2N ¼ 86) (2N ¼ 84) (nominal) (95% CI)

*0102 7.0 3.6 (0.50) *0301 24.2 9.4 0.004 3.07 2.3 16.7 0.01 0.12 (1.62–5.81) (0.01–0.55) *0403 9.1 8.3 (0.81) 9.3 7.1 (0.61) *0404 4.7 7.1 (0.53) *0701 3.8 9.4 (0.05) 5.8 6.0 (0.97) *1101 6.1 10.0 (0.21) 10.5 13.1 (0.59) *1104 8.3 9.4 (0.73) 17.4 16.7 (0.89) *1303 8.3 6.1 (0.45) *1401 3.8 6.7 (0.27) *1501 9.1 4.4 (0.10) 16.3 6.0 (0.03) 3.07 (1.05–8.96) *1502 1.5 5.6 (0.08) Othersb 25.8 30.6 (0.35) 26.7 23.8 (0.66) Global Pc 0.005 0.047 aValues are presented after the Bonferroni correction (corrected for 10 tests in Muslims, 9 tests in Christians), or as nominal P-values in parentheses. P, OR and CI values shown are from Pearson’s w2-tests except for the DRB1*0301 allele in Christian Arabs, which was computed by Fisher’s exact test. bOthers include the following DRB1 alleles that were less than 5% frequent—Muslims: DRB1*-0101, 0102, 0302, 0401, 0402, 0404, 0405, 0406, 0801, 0804, 0806, 0901, 1001, 1102, 1103, 1201, 1301, 1302, 1305, 1503, 1601 and 1602; Christians: DRB1*-0101, 0401, 0402, 0405, 0801, 0804, 0901, 1001, 1103, 1201, 1302, 1303, 1401, 1402, 1502 and 1602. cGlobal Pearson’s w2-test and w2-exact test for case–control analysis in Muslims and Christians, respectively.

Table 4 HLA DRB1*0301 genotype association tests

DRB1 genotype Muslims Christians

Cases (N ¼ 66) Controls (N ¼ 90) Cases (N ¼ 43) Controls (N ¼ 42)

*0301/*0301 6003 *0301/other 20 17 2 11 Other/other 40 73 41 28 P 0.002 0.02

Table 5 HLA-DQB1 percent allele frequencies

DQB1 Muslims Christians

a a Cases Controls PB OR Cases Controls PB OR (2N ¼ 132) (2N ¼ 180) (nominal) (95% CI) (2N ¼ 86) (2N ¼ 84) (nominal) (95% CI)

*0201 24.2 9.4 0.004 3.07 2.3 16.7 0.01 0.12 (1.62–5.81) (0.01–0.55) *0202 5.3 10.6 (0.10) 8.1 8.3 (0.96) *0301 21.2 29.4 (0.10) 38.4 34.5 (0.60) *0302 11.4 13.3 (0.60) 11.6 9.5 (0.66) *0402 8.1 7.1 (0.81) *0501 12.9 5.0 (0.01) 2.81 9.3 7.1 (0.61) (1.21–6.52) *0503 3.0 6.7 (0.20) *0601 2.3 6.1 (0.16) *0602 6.8 2.2 (0.08) 12.8 4.8 (0.10) Othersb 12.9 17.2 (0.29) 9.3 11.9 (0.58) Global Pc o0.001 0.07 aValues are presented after the Bonferroni correction (corrected for 9 tests in Muslims, 8 tests in Christians), or as nominal P-values in parentheses. P, OR and CI values shown are from Pearson’s w2-tests except for the DQB1*0201 allele in Christian Arabs, which was computed by Fisher’s exact test. bOthers include the following DQB1alleles that were less than 5% frequent—Muslims: DQB1*-0303, 0305, 0402, 0502, 0603, 0604 and 0609; Christians: DQB1*-0303, 0305, 0502, 0503, 0601, 0603 and 0604. cGlobal Pearson’s w2-test for case–control analysis.

Genes and Immunity Genetics of multiple sclerosis in Israeli Arabs G Benedek et al 426 Table 6 HLA-B percent allele frequencies

B Muslims Christians

a Cases Controls PB OR Cases Controls P (nominal) (2N ¼ 132) (2N ¼ 178) (nominal) (95% CI) (2N ¼ 86) (2N ¼ 84)

*07 5.3 2.8 (0.25) *08 9.9 3.9 (0.03) 2.7 (1.05–7.0) *14 3.5 6.0 (0.49) *15 7.6 4.5 (0.24) 5.8 4.8 (0.76) *18 3.0 5.1 (0.57) 8.1 10.7 (0.57) *35 15.9 9.6 (0.08) 17.4 10.7 (0.21) *38 1.5 5.6 (0.08) *41 6.8 9.0 (0.51) 8.1 8.3 (0.96) *44 7.6 6.2 (0.61) 12.8 9.5 (0.50) *49 7.6 5.6 (0.47) 10.5 7.1 (0.45) *50 7.6 6.7 (0.76) *51 3.8 8.4 (0.11) 7.0 13.1 (0.18) *52 0.0 7.3 0.01 0.03 (0.02–0.23) Othersb 23.5 25.3 (0.81) 26.7 29.8 (0.66) Global Pc 0.009 0.77

aValues are presented after the Bonferroni correction (corrected for 13 tests), or as nominal P-values in parentheses. P, OR and CI values shown are from Pearson’s w2-tests except for the HLA-B*52 allele in Muslims, which was computed by Fisher’s exact test. bOthers include the following HLA-B alleles that were less than 5% frequent—Muslims: HLA B*-13, 14, 27, 37, 39, 40, 42, 45, 47, 53, 55, 56, 57, 58 and 73; Christians: HLA B*-07, 08, 13, 27, 37, 38, 40, 45, 47, 50, 52, 53, 55, 57, 58 and 73. cGlobal Pearson’s w2-test and w2 exact test for case–control analysis in Muslims and Christians, respectively.

The DRB1*1501 allele was nominally associated with MS Table 7 HLA-A and -C percent allele frequencies only in the Christian Arab group (P ¼ 0.033, OR ¼ 3.07); HLA Muslims Christians however, this association did not pass correction for multiple testing, likely due to the small cohort size. A Cases Controls Cases Controls Testing for genotype association supports a significant (2N ¼ 132) (2N ¼ 180) (2N ¼ 86) (2N ¼ 84) effect for DRB1*0301 genotypes in both Muslims and Christian Arabs (P ¼ 0.002 and 0.02, respectively). *01 15.9 16.7 17.4 11.9 Analysis of the DQB1 locus revealed a significant *02 19.7 27.2 12.8 14.3 difference between cases and controls in the Muslim *03 10.6 7.2 9.3 17.9 (P ¼ 0.0002), and trend for the Christian population *11 5.6 5.6 (P ¼ 0.065) (Table 5). The DQB1*0201 allele, which is part *24 12.9 10.0 14.0 19.1 of the known DRB1*0301-DQB1*0201 haplotype, was, as *30 9.9 6.1 5.8 9.6 *32 11.6 4.8 expected, associated with the disease in both ethnicities *68 7.0 3.6 although in different directions. In Muslims, a nominal Othersa 27.3 27.2 22.1 19.1 association was observed also for the DQB1*0501 Global Pb 0.51 0.31 (P ¼ 0.013, OR ¼ 2.81). Haplotype analysis indicated that the DQB1*0501 allele was present on several haplotypes Cw 2N ¼ 132 2N ¼ 180 2N ¼ 86 2N ¼ 84 with different HLA-DRB1 alleles (Table 8); however, the *03 6.8 4.4 1.2 8.3 *04 19.7 11.7 22.1 14.3 small sample size of this study precludes any definitive *06 18.2 13.3 8.1 7.1 conclusions regarding the role of this allele and its *07 23.5 19.4 22.1 19.1 associated haplotypes. *08 3.5 6.0 After adjusting for gender effect as a confounding *12 7.6 16.6 20.9 15.5 variable, logistic regression modeling supported a major *15 4.5 6.1 4.7 9.5 genetic effect from the DRB1*0301 in both the Muslim *16 5.3 7.9 5.8 3.6 ¼ ¼ ¼ *17 6.1 8.3 7.0 8.3 (P 0.0008, OR 3.17) and the Christian Arab (P 0.004, Othersc 8.3 12.2 4.7 8.3 OR ¼ 0.10) cohorts, with exclusion of the DQB1*0201 Global Pb 0.10 0.32 allele from the model as a dependent variable. The allelic logistic association was significant for the DQB1*0501 aOthers include the following HLA-A alleles that were less than 5% allele (P ¼ 0.03) as well in the Muslim cohort. frequent—Muslims: HLA-A*-23, 26, 29, 31, 32, 33, 34, 36, 66, 68, 69, The contradictory association signal observed in 74 and 80; Christians: HLA-A*-11, 23, 26, 29, 31, 33, 69 and 80. Muslim and Christian Arabs for the DRB1*0301- bGlobal Pearson’s w2-test and w2 exact test for case–control analysis, DQB1*0201 haplotype might be due to a genetic in Muslims and Christians, respectively. difference within or beyond the HLA region. We there- cOthers include the following HLA-C alleles that were less than 5% fore analyzed 21 single nucleotide polymorphisms frequent—Muslims: HLA-C*-01, 02, 05, 08, 14 and 18; Christians: (SNPs) across the major histocompatibility complex HLA-C*-01, 02, 05 and 14. region in a subgroup of cases and controls of both

Genes and Immunity Genetics of multiple sclerosis in Israeli Arabs G Benedek et al 427 7 7 Religious affiliation was used as a proxy for popula- tion classification. This decision was prompted by the observation of different rates of MS in Muslim and rs5020946 rs215733 rs660895 rs7451962 rs312975 rs3129763 rs9272346 rs9272535 rs9272723 rs3891175 Muslims Christians Christian Israeli Arabs.19 Moreover, although Christian- C..A..A..A..C..A..A..A..C..A 97% (36) 93% (14) DRB1*0301 ity and Islam have coexisted for many centuries in the C..A..A..A..C..A..A..A..C..G 3% (1) 7% (1) Middle East, religion has effectively been a barrier to population admixture. Genetic studies reporting a strong 75% (9) C..A..A..G..A..G..G..G..T..G 53% (8) effect for religion on Y-chromosome marker distributions C..G..A..G..A..G..G..G..T..G 17% (2) 47% (7) DRB1*1501 in the neighboring Lebanese population,21 and the C..A..A..G..C..G..G..G..T..G 8% (1) 0% different allele frequency distribution observed herein for five HLA class I and class II genes, support analyzing the data stratified by religion. It is of interest that allele telomere centromere frequency differences were observed mainly between DRA DRB5DRB1 DQA1 DQB1 DQA2 case groups, suggesting that different genetic factors, govern, at least in part, disease risk in the two religion Figure 1 Conservation of the DRB1*0301-associated SNP haplo- type between Muslim and Christian Arabs. The 10-SNP haplotypes groups. inferred across the DRB1-DQB1 region are shown for their corresponding DRB1*0301 or DRB1*1501 alleles. Haplotype dis- Clinical characteristics of patient cohorts tribution in percentages in Muslims and Christian Arabs is shown The Muslim and Christian Arab patient cohorts in this left to the nucleotide sequence, and the number of observations is study appear to share similar demographics and disease presented in parentheses. rs numbers for each SNP (above characteristics. Clinical characteristics were similar to haplotype sequence) and corresponding location in the DR–DQ 22 region (below haplotype sequence) are marked according to the that described in European populations, African- 23,24 UCSC genome browser March 2006 (hg 18) http://genome.ucsc.edu/. American populations as well as other Arab popula- tions.16–18,25,26 At the time of recruitment, only a minority of the patients was at an advanced stage of disease. This subethnicities and imputed parental haplotypes. For the may reflect the relative young age of the cohorts, the DRB1*0301-DQB1*0201 haplotype, a highly conserved modification of the natural history of the disease by the and specific 10-SNP haplotype was present across the immunomodulatory treatments, as well as an inherent HLA-DRB1 and -DQB1 gene regions (Figure 1), in both recruitment bias for less disabled MS individuals, Muslim and Christian Arab individuals, suggesting their capable of attending the MS clinics and cognitively genetic background across this locus is similar and capable of informed consent. Primary progressive course cannot therefore account for the opposing associations of disease rates are lower than reported for populations observed. The SNP haplotype was specifically associated of European descent (B15%),22,24 but similar to rates with DRB1*0301-DQB1*0201, as it was present in our reported in other Arab populations (3.6–7.9%).16–18 data set only once (out of 51 chromosomes) on the However, the small cohort size and the recruitment bias background of another DRB1-DQB1 haplotype. The may account for the low primary progressive rates. conservation between Muslim and Christian Arab SNP haplotypes was not a general phenomenon, as Allele distribution of HLA genes in Israeli Arabs DRB1*1501-associated haplotypes showed higher hetero- Comparison of the HLA data in the control populations geneity. of this study with published data from other Arab cohorts, as well as Y chromosome genetic studies, HLA class I analyses emphasizes the presence of regional population stratifi- 27–30 Global case–control analysis for HLA class I genes (HLA- cation. These observations highlight the importance A,-B,-C; Tables 6 and 7) was significant only for HLA-B of procuring a locality- and religion-based set of matched locus in the Muslim population (global P ¼ 0.009). The controls for association studies in Arab populations, as HLA-B*52 allele remained significant after correcting for was carried out for this study. multiple testing, was negatively associated with MS The few studies available on classical HLA genes in ¼ ¼ MS patients of Arab descent are based on serological (PB 0.01, OR 0.03) and totally absent from the patient 31–34 population. HLA-B*52 also emerged as an independent testing. These studies report association of the DR2 factor from gender and other HLA class II alleles in allelic family (that includes also the DRB1*15 allele logistic regression modeling (P ¼ 0.04). group) and DR3 with MS. Thus, this study is the first to provide detailed high-resolution data for the HLA-A,-B, -C, DRB1 and -DQB1 allele distribution in Christian Discussion Arabs and association with MS risk in populations of Arab descent. In this study, we investigated the associations between HLA class I and class II genes and MS in the Israeli Arab DRB1*0301-DQB1*0201 haplotype and MS risk in Israeli population, and for the first time present allele frequency Arabs data for the Christian Arab population. The HLA region, The most interesting result of this study is the opposite and particularly the class II HLA-DRB1 gene locus, has associations for the DRB1*0301-DQB1*0201 haplotype in been shown in numerous studies to be associated with Muslim versus Christian Arabs as deduced from the MS susceptibility. The results of this study confirm the DRB1*0301 and DQB1*0201 allelic tests. The increased genetic effect of this region in Israeli Arab populations, MS risk associated with this haplotype has been while highlighting the complex contributions of the previously reported in both populations at high and HLA-DRB1*0301 allele and its extended haplotype. low risk for MS.7,12,35,36 This is the first report that may

Genes and Immunity Genetics of multiple sclerosis in Israeli Arabs G Benedek et al 428 indicate a ‘protective’ effect for the same haplotype, in a Table 8 DRB1-DQB1 haplotypes based on familial segregation, population previously unstudied, the Christian Arabs. A homozygote and inference analyses similar phenomenon was reported for DRB1*07, which is negatively associated with type I diabetes (T1D) in Selected allele Allele in linkagea Linkage frequency Yemenite Jews and a few other populations, but is positively associated with type I diabetes in African Muslims Christians blacks.37 Studies in humanized mouse harboring transgenic DRB1*0301 DQB1*0201 17/17 5/5 DRB1*0301 suggest that for this allele genetic interactions DRB1*1501 DQB1*0501 5/6 DQB1*0602 1/6 6/6 in trans have a major role in determining disease non DQB1*0501b or 3/18 4/19 outcome in the experimental autoimmune encephalitis DQB1*0602 model. Thus, presence of DQB1*0601 protected the mice DQB1*0501 DRB1*0101 3/16 1/6 from experimental autoimmune encephalitis whereas DRB1*0102 3/16 5/6 presence of DQB1*0302 led to higher disease incidence DRB1*1001 4/16 and severity.38,39 DRB1*1101 1/16 The embedded SNP haplotype for the DRB1*0301- DRB1*1501 5/16 DQB1*0201 appeared to be conserved across Muslim and a Christian individuals, implying that its contrasting Analyses based on familial or homozygote segregation of alleles, except where noted otherwise. genetic effects may result from a modifier outside the b class II region, exerting epistasis effects. Although Inference from all carriers of the DRB1*1501 allele. Identity of allele Muslim and Christian Arabs in Israel share a similar in linkage could not be determined due to small sample size. geographical environment, it should be noted that lifestyle differences are present (including dietary differ- ences, for example, as dictated by religion, family in light of the recent suggestion for a protective role in size and so on), which may also confer a modifying MS for the HLA-Bw4 serological group, as the HLA-B*52 effect through gene–environment interactions.40 Alterna- carries this serological specificity,41 possibly relating to tively, low frequency or rare genetic polymorphisms HLA-Bw4 function as a ligand for the killer immunoglo- within the DRB1*0301-DQB1*0201 haplotype block, bulin-like receptors. Notably, HLA-B*52 is more preva- which are not captured by the SNPs studied herein, lent in world regions with low MS prevalence such as remain a possibility. Japan,10,42 where its frequency was reported as lower in The negative association of the DRB1*0301 haplotype MS patients, although not statistically significant.43 with MS in Christian Arabs is based on its low frequency among recruited cases. It is unlikely that this observation Conclusions was derived from inclusion of related individuals in the This study confirms the importance of the HLA-DRB1 data set, as our recruitment strategy included meticulous gene for MS susceptibility in Arab populations, and assessment of possible relatedness between participants provides data on MS in the Christian Arab population through assessment of surnames and clans. Moreover, for the first time. By using religion as a population the participants of the control group were recruited, classifier, we were able to show significant though where possible, from the same localities as the cases, to contrasting effects for the DRB1*0301-DQB1*0201 haplo- ensure a similar representation of the regional genetic type in these populations. Therefore, the overall effect of background. Therefore, the observed lower frequency of the DRB1*0301-DQB1*0201 haplotype on MS risk is the DRB1*0301 haplotype in affected Christian Arabs is likely determined through interactions with additional likely to be a true observation. genetic and environmental factors. Future studies on these Arab populations in Israel, the neighboring Other HLA gene associations and MS susceptibility in the countries and immigration countries with congregations Israeli Arab populations of these populations, such as the United States, will allow DRB1*1501 is considered the HLA allele with the 1 expansion of these findings on genetic effects of the HLA strongest genetic effect known on MS. The frequency genes, and possibly others, to MS susceptibility and of the DRB1*1501 allele in the Israeli Arab population is disease course. low, limiting the power of the study to detect effects of this allele. Accordingly, although the DRB1*1501 was more frequent among cases, nominal significance was Materials and methods observed only in the Christian cohort. Haplotype analysis based on family data, homo- Study participants zygotes and inference suggests that in Israeli Arabs, in Participants (418 years), who self-identified as Christian addition to the European haplotype DRB1*1501- or Muslim Arabs, were recruited at three MS referral DQB1*0602, the DRB1*1501 is present also on haplotypes centers in Israel (Carmel Medical Center, Haifa; Hadassah in linkage with DQB1*0501 and others (Table 8). Medical Center, Jerusalem and Sieff Medical Center, This suggests the haplotype structure in Israeli Arab ). Participants included patients with clinically populations differs from that prevalent in Europe, definite or laboratory supported MS diagnosis according confirming their advantage for genetic studies of the to Poser and MacDonald criteria,44,45 their parents or HLA genes in MS. siblings, where available, and healthy individuals as The HLA-B*52-negative association we observed in the controls. Only one affected family member was included Muslim cohort is a new observation that remains to be from multicase families (seven families). A total of 109 further validated. This finding is even more interesting (43 Christians and 66 Muslims) MS patients and 132 (42

Genes and Immunity Genetics of multiple sclerosis in Israeli Arabs G Benedek et al 429 Christians and 90 Muslims) healthy controls were Statistical analysis included in the final analysis. On the basis of our earlier Statistical analysis of clinical and demographic data was epidemiological survey,19 we estimate this cohort to conducted using SPSS version 15.0 (SPSS Inc., Chicago, represent about 60% of the Muslim Israeli Arab MS IL, USA). Categorical variables were compared using patients and 70% of the Christian Arab patients. The w2-test. Continuous variables were tested for normality study protocol received ethical approval from the Israeli by the nonparametric one sample Kolmogorov–Smirnov Ministry of Health Helsinki Committee for Genetics test, and then compared using t-tests or Mann–Whitney Studies. All participants signed an informed consent. tests, as appropriate. Unrelated controls were enrolled mainly from spouses Allelic frequencies of the HLA-DRB1,-DQB1, HLA-A, and friends of patients, with emphasis on matching HLA-B and HLA-C genes were compared using SPSS residence where possible; exclusion criteria included version 15.0 and JMP-Genomics version 3.2 software presence of MS in family members up to third degree, or applications (Cary, NC, USA). Global w2-tests and exact presence of any autoimmune or chronic inflammatory tests using Monte Carlo repeated sampling methods condition. Individuals were included in the genetic (10 000 sampled tables) were used to compare allele analysis only if all four grandparents were of the same distributions between Muslims and Christians for each subethnic affiliation to minimize stratification effects. HLA gene in case and control groups. Case–control Detailed demographic and clinical data were obtained comparisons were then performed within each ethnic from all subjects. MS disease type (relapsing-remitting, group using the Pearson’s w2-test or exact tests using secondary-progressive, relapsing-progressive or pri- Monte Carlo methods as appropriate, if a cell with a mary-progressive) and Expanded Disability Status Scale count of less than 5 was present. Rare alleles with a (EDSS46) score were recorded at date of clinical report. frequency of o5% in both cases and controls groups DNA was extracted from blood samples using stan- were pooled. For specific alleles 2 Â 2 Pearson’s w2-test or dard procedures (Puregene, Gentra Systems, Minneapo- Fisher’s exact test were computed. P-values lower than lis, MN, USA), or from saliva samples using the Oragene 0.05 were considered statistically significant after apply-

DNA kit (DNA-Genotek, Ottawa, Ontario, Canada). ing the Bonferroni correction (PB). The OR and 95% CI were calculated for statistically significant alleles, using HLA typing asymptotic confidence limits, or exact confidence limits HLA-DRB1,-DQB1,-A and -B loci at low resolution were when Fisher’s exact test was used, or the ‘median typed using the Lipa HLA kit (Murex Innogenetics, unbiased estimation’ for OR and CI when empty cells Ghent, Belgium). were present.47 Deviation from Hardy–Weinberg equili- High-resolution typing of the HLA-DRB1*03,*11,*13 brium was tested in controls, for each allele versus all and *15 alleles used the PCR sequence-specific primer other alleles as a pool. Genotype association test was method (Olerup kit, Saltsjobaden, Sweden). High-resolu- applied for the DRB1*0301 allele, implementing the exact tion typing of the remaining alleles was performed with Monte Carlo estimation for the P-value (10 000 repeated the Lipa HLA-DRB decoder kit (Murex Innogenetics, samplings) to account for the small sample size available. Gent, Belgium). A forward stepwise conditional logistic regression HLA-C was typed using the LIFECODES HLA-C analysis was performed with possible confounding typing kit (Tepnel, Lifecodes, Stamford, CT, USA). The variables, such as gender, in the model. The criteria for samples were analyzed with the Luminex 100 system stepwise selection were set at entrance P-value p0.05 (Luminex, Austin, TX, USA). Genotyping was performed and removal P-value X0.15. All P-values reported (Wald masked to the status of the sample. test) are two sided. All samples were successfully genotyped except for SNP haplotypes were inferred using the EM algo- one sample that failed HLA-B typing. Available family rithms as implemented in JMP Genomics, only inferred members were typed for haplotype reconstruction, as haplotypes with a probability of 1 were accepted. HLA well as to check for Mendelian inconsistencies. haplotypes were inferred, where possible, by using family data or as derived from homozygote data. SNP genotyping Illumina GoldenGate MHC mapping panel GT-17-180 SNP data (1293 SNPs) were available for 68% of the Conflict of interest Muslim and 91% of the Christian data sets. Data filtering removed samples and SNPs with success rate o95%. The authors declare no conflict of interest. Owing to small cohort sizes, stringent criteria were used excluding SNPs with minor allele frequency o10%; or Hardy–Weinberg equilibrium test at Pp0.01. The final Acknowledgements number of SNPs available for analysis was 937. Twenty- one SNPs (rs2524005, rs2860580, rs3094141, rs2853961, This study was supported by the National Multiple rs4361609, rs3873386, rs2156875, rs2523586, rs2523567, Sclerosis Society of USA–RG 3520A1/1 (AM); the rs2844580, rs2844575, rs2253908, rs5020946, rs660895, Wolfson Family Charitable Trust (AM), the Sacta-Rashi rs7451962, rs6931337, rs3129763, rs9272346, rs9272535, Foundation (AM), the Technion’s Vice-President for rs9272723 and rs3891175) were manually selected to span Research funds (AM) and by Grant AI067152 from the the HLA class I and class II regions, taking into account National Institutes of Allergy and Infectious Diseases inter- or intragenic location and relative spacing between (JRO). We thank the MS patients, their families and the SNPs. The final data set included 52 affected and 54 healthy volunteers for participation in this study. We controls in the Muslim cohort; 40 affected and 37 controls thank Yvonne Nachleh, Anat Volkowich, Matana Galili from the Christian cohort. and Maayan Merhav, Rappaport Faculty of Medicine and

Genes and Immunity Genetics of multiple sclerosis in Israeli Arabs G Benedek et al 430 Research Institute, Technion–Israel Institute of Technol- 17 Yamout B, Barada W, Tohme RA, Mehio-Sibai A, Khalifeh R, ogy, Haifa, Israel, for expert technical assistance; Sara El-Hajj T. Clinical characteristics of multiple sclerosis in Dishon, Carmel Medical Center, Haifa, Israel, for Lebanon. J Neurol Sci 2008; 270: 88–93. providing patient care and assistance in clinical data 18 El-Salem K, Al-Shimmery E, Horany K, Al-Refai A, Al-Hayk management; Dr Ofra Barnett-Griness and Nili Stein K, Khader Y. Multiple sclerosis in Jordan: a clinical and Carmel Medical Center, Haifa, Israel, for statistical epidemiological study. J Neurol 2006; 253: 1210–1216. consultations and Prof Doron Lancet, Weizmann Insti- 19 Alter M, Kahana E, Zilber N, Miller A. Multiple sclerosis 66 tute of Science, for critical reviewing of the article. frequency in Israel’s diverse populations. Neurology 2006; : 1061–1066. 20 Shifman S, Darvasi A. The value of isolated populations. Nat Genet 2001; 28: 309–310. 21 Zalloua PA, Xue Y, Khalife J, Makhoul N, Debiane L, Platt DE References et al. Y-chromosomal diversity in Lebanon is structured by recent historical events. Am J Hum Genet 2008; 82: 873–882. 1 Oksenberg JR, Baranzini SE, Sawcer S, Hauser SL. The 22 Confavreux C, Vukusic S, Adeleine P. Early clinical genetics of multiple sclerosis: SNPs to pathways to patho- predictors and progression of irreversible disability in genesis. Nat Rev Genet 2008; 9: 516–526. multiple sclerosis: an amnesic process. Brain 2003; 126 2 Ramagopalan SV, Deluca GC, Degenhardt A, Ebers GC. The (Part 4): 770–782. genetics of clinical outcome in multiple sclerosis. J Neuro- 23 Cree BA, Khan O, Bourdette D, Goodin DS, Cohen JA, Marrie immunol 2008; 201–202: 183–199. RA et al. Clinical characteristics of African Americans vs 3 Yeo TW, De Jager PL, Gregory SG, Barcellos LF, Walton A, Caucasian Americans with multiple sclerosis. Neurology 2004; Goris A et al. A second major histocompatibility complex 63: 2039–2045. susceptibility locus for multiple sclerosis. Ann Neurol 2007; 61: 24 Marrie RA, Cutter G, Tyry T, Vollmer T, Campagnolo D. Does 228–236. 4 Fogdell-Hahn A, Ligers A, Gronning M, Hillert J, Olerup O. multiple sclerosis-associated disability differ between races? 66 Multiple sclerosis: a modifying influence of HLA class I genes Neurology 2006; : 1235–1240. in an HLA class II associated autoimmune disease. Tissue 25 Daif AK, Al-Rajeh S, Awada A, Al Bunyan M, Ogunniyi A, Antigens 2000; 55: 140–148. AbdulJabar M et al. Pattern of presentation of multiple 5 Chao MJ, Barnardo MC, Lincoln MR, Ramagopalan SV, sclerosis in Saudi Arabia: analysis based on clinical and Herrera BM, Dyment DA et al. HLA class I alleles tag HLA- paraclinical features. Eur Neurol 1998; 39: 182–186. DRB1*1501 haplotypes for differential risk in multiple 26 Al-Din AS, Khogali M, Poser CM, al-Nassar KE, Shakir R, sclerosis susceptibility. Proc Natl Acad Sci USA 2008; 105: Hussain J et al. Epidemiology of multiple sclerosis in Arabs in 13069–13074. Kuwait: a comparative study between Kuwaitis and Palesti- 6 Rioux JD, Goyette P, Vyse TJ, Hammarstrom L, Fernando MM, nians. J Neurol Sci 1990; 100: 137–141. Green T et al. Mapping of multiple susceptibility variants 27 Almawi WY, Busson M, Tamim H, Al-Harbi EM, Finan RR, within the MHC region for 7 immune-mediated diseases. Proc Wakim-Ghorayeb SF et al. HLA class II profile and distribution Natl Acad Sci USA 2009; 106: 18680–18685. of HLA-DRB1 and HLA-DQB1 alleles and haplotypes among 7 Oksenberg JR, Barcellos LF, Cree BA, Baranzini SE, Bugawan Lebanese and Bahraini Arabs. Clin Diagn Lab Immunol 2004; 11: TL, Khan O et al. Mapping multiple sclerosis susceptibility to 770–774. the HLA-DR locus in African Americans. Am J Hum Genet 28 Amar A, Kwon OJ, Motro U, Witt CS, Bonne-Tamir B, Gabison 2004; 74: 160–167. R et al. Molecular analysis of HLA class II polymorphisms 8 Caillier SJ, Briggs F, Cree BA, Baranzini SE, Fernandez-Vina among different ethnic groups in Israel. Hum Immunol 1999; M, Ramsay PP et al. Uncoupling the roles of HLA-DRB1 and 60: 723–730. HLA-DRB5 genes in multiple sclerosis. J Immunol 2008; 181: 29 Sanchez-Velasco P, Karadsheh NS, Garcia-Martin A, Ruiz de 5473–5480. Alegria C, Leyva-Cobian F. Molecular analysis of HLA allelic 9 Cree BA, Reich DE, Khan O, De Jager PL, Nakashima I, frequencies and haplotypes in Jordanians and comparison with Takahashi T et al. Modification of multiple sclerosis pheno- other related populations. Hum Immunol 2001; 62: 901–909. types by African ancestry at HLA. Arch Neurol 2009; 66: 226–233. 30 Nebel A, Filon D, Weiss DA, Weale M, Faerman M, 10 Pugliatti M, Sotgiu S, Rosati G. The worldwide prevalence of Oppenheim A et al. High-resolution Y chromosome haplo- multiple sclerosis. Clin Neurol Neurosurg 2002; 104: 182–191. types of Israeli and Palestinian Arabs reveal geographic 11 Marrosu MG, Murru MR, Costa G, Cucca F, Sotgiu S, Rosati G substructure and substantial overlap with haplotypes of Jews. et al. Multiple sclerosis in Sardinia is associated and in linkage Hum Genet 2000; 107: 630–641. disequilibrium with HLA-DR3 and -DR4 alleles. Am J Hum 31 Al-Din AS, Al-Saffar M, Siboo R, Behbehani K. Association Genet 1997; 61: 454–457. between HLA-D region epitopes and multiple sclerosis in 12 Marrosu MG, Murru MR, Costa G, Murru R, Muntoni F, Arabs. Tissue Antigens 1986; 27: 196–200. Cucca F. DRB1-DQA1-DQB1 loci and multiple sclerosis 32 Al-Din AS, el-Khateeb M, Kurdi A, Mubaidin A, Wriekat A, predisposition in the Sardinian population. Hum Mol Genet al-Shehab A et al. Multiple sclerosis in Arabs in Jordan. 1998; 7: 1235–1237. J Neurol Sci 1995; 131: 144–149. 13 Grossman I, Avidan N, Singer C, Paperna T, Lancet D, 33 Al-Din ASN, Kurdi A, Mubaidin A, El-Khateeb M, Khalil RW, Beckmann JS et al. Genomic profiling of interpopulation Wriekat AL. Epidemiology of multiple sclerosis in Arabs in diversity guides prioritization of candidate-genes for auto- Jordan: a comparative study between Jordanians and Pales- immunity. Genes Immun 2004; 5: 493–504. tinians. J Neurol Sci 1996; 135: 162–167. 14 Brassat D, Salemi G, Barcellos LF, McNeill G, Proia P, Hauser 34 Al-Shammri S, Nelson RF, Al-Muzairi I, Akanji AO. HLA SL et al. The HLA locus and multiple sclerosis in Sicily. determinants of susceptibility to multiple sclerosis in an Neurology 2005; 64: 361–363. Arabian Gulf population. Mult Scler 2004; 10: 381–386. 15 Kallio SP, Jakkula E, Purcell S, Suvela M, Koivisto K, Tienari PJ 35 Barcellos LF, Sawcer S, Ramsay PP, Baranzini SE, Thomson G, et al. Use of a genetic isolate to identify rare disease variants: Briggs F et al. Heterogeneity at the HLA-DRB1 locus and C7 on 5p associated with MS. Hum Mol Genet 2009; 18: risk for multiple sclerosis. Hum Mol Genet 2006; 15: 2813–2824. 1670–1683. 36 Stankovich J, Butzkueven H, Marriott M, Chapman C, Tubridy 16 Alshubaili AF, Alramzy K, Ayyad YM, Gerish Y. Epidemiol- N, Tait BD et al. HLA–DRB1 associations with disease ogy of multiple sclerosis in Kuwait: new trends in incidence susceptibility and clinical course in Australians with multiple and prevalence. Eur Neurol 2005; 53: 125–131. sclerosis. Tissue Antigens 2009; 74: 17–21.

Genes and Immunity Genetics of multiple sclerosis in Israeli Arabs G Benedek et al 431 37 Israel S, Kwon OJ, Weintrob N, Sprecher E, Bloch K, Assa S 42 Middleton D, Menchaca L, Rood H, Komerofsky R. New allele et al. HLA class II immunogenetics of IDDM in Yemenite Jews. frequency database: http://www.allelefrequencies.net Tissue Hum Immunol 1998; 59: 728–733. Antigens 2003; 61: 403–407. 38 Mangalam A, Luckey D, Basal E, Behrens M, Rodriguez M, 43 Ono T, Zambenedetti MR, Yamasaki K, Kawano Y, Kamika- David C. HLA-DQ6 (DQB1*0601)-restricted T cells waji N, Ito H et al. Molecular analysis of HLA class I (HLA-A protect against experimental autoimmune encephalomyelitis and -B) and HLA class II (HLA-DRB1) genes in Japanese in HLA-DR3.DQ6 double-transgenic mice by generating patients with multiple sclerosis (Western type and Asian anti-inflammatory IFN-gamma. J Immunol 2008; 180: type). Tissue Antigens 1998; 52: 539–542. 7747–7756. 44 McDonald WI, Compston A, Edan G, Goodkin D, Hartung 39 Mangalam A, Luckey D, Basal E, Jackson M, Smart M, HP, Lublin FD et al. Recommended diagnostic criteria for Rodriguez M et al. HLA-DQ8 (DQB1*0302)-restricted multiple sclerosis: guidelines from the International Panel on the Th17 cells exacerbate experimental autoimmune encephalo- Diagnosis of Multiple Sclerosis. Ann Neurol 2001; 50: 121–127. myelitis in HLA-DR3-transgenic mice. J Immunol 2009; 182: 45 Poser CM, Paty DW, Scheinberg L, McDonald WI, Davis FA, 5131–5139. Ebers GC et al. New diagnostic criteria for multiple sclerosis: 40 Kivity S, Agmon-Levin N, Blank M, Shoenfeld Y. Infections guidelines for research protocols. Ann Neurol 1983; 13: 227–231. and autoimmunity—friends or foes? Trends Immunol 2009; 30: 46 Kurtzke JF. Rating neurologic impairment in multiple sclero- 409–414. sis: an Expanded Disability Status Scale (EDSS). Neurology 41 Lorentzen AR, Karlsen TH, Olsson M, Smestad C, Mero IL, 1983; 33: 1444–1452. Woldseth B et al. Killer immunoglobulin-like receptor ligand 47 Parzen M, Lipsitz S, Ibrahim J, Klar N. An estimate of the HLA-Bw4 protects against multiple sclerosis. Ann Neurol 2009; odds ratio that always exists. J Comput Graph Stat 2002; 11: 65: 658–666. 420–436.

Appendix Members of The Israeli MS Genetics Group:

Ester Kahana, MD: Barzilai Medical Center, ; Center; Panayiota Petrou, MD: Hadassah Medical Cen- Dimitrios Karussis, MD PhD: Hadassah Medical Center, ter, Jerusalem; Hana Rawasdeh, MD: Carmel Medical Jerusalem; Rafik Masalha, MD: , Center, Haifa; Radi Shahein, MD: Sieff Medical Center, Beer-Sheva; Ariel Miller, MD PhD: Carmel Medical Safed.

Genes and Immunity