A Cytokine Gene Screen Uncovers SOCS1 As Genetic Risk Factor for Multiple Sclerosis

ORIGINAL ARTICLE A cytokine gene screen uncovers SOCS1 as genetic risk factor for multiple sclerosis

K Vandenbroeck1,2, J Alvarez3, B Swaminathan1, I Alloza1, F Matesanz4, E Urcelay5, M Comabella6, A Alcina4, M Fedetz4, MA Ortiz5, G Izquierdo7, O Fernandez8, N Rodriguez-Ezpeleta3, C Matute9, S Caillier10, R Arroyo5, X Montalban6, JR Oksenberg10, A Antigu¨ edad11 and A Aransay3 1Neurogenomiks Laboratory, Department of Neuroscience, University of the Basque Country UPV/EHU, Leioa, Spain; 2IKERBASQUE, Basque Foundation for Science, Bilbao, Spain; 3CIC bioGUNE, Parque Tecnolo´gico de Bizkaia, Derio, Spain; 4Instituto de Parasitologıáy Biomedicina ‘Lo´pez Neyra’, Consejo Superior de Investigaciones Cientı´ficas, Granada, Spain; 5Immunology and Neurology Department, Hospital Clıńico S Carlos, Instituto de Investigacioń Sanitaria del Hospital Clıńico San Carlos (IdISSC), Madrid, Spain; 6Centre d’Esclerosi Mu´ltiple de Catalunya, CEM-Cat, Unitat de Neuroimmunologia Clıńica, Hospital Universitari Vall d’Hebron, Barcelona, Spain; 7Unidad de Esclerosis Mu´ltiple, Hospital Virgen Macarena, Sevilla, Spain; 8Servicio de Neurologıá, Instituto de Neurociencias Clıńicas del Hospital Regional Universitario Carlos Haya de Ma´laga, Ma´laga, Spain; 9Neurotek Laboratory, Department of Neuroscience, University of the Basque Country UPV/EHU, Leioa, Spain; 10Department of Neurology, University of California, San Francisco, CA, USA and 11Servicio de Neurologıá, Hospital de Basurto, Bilbao, Spain

Cytokine and cytokine receptor genes, including IL2RA, IL7R and IL12A, are known risk factors for multiple sclerosis (MS). Excitotoxic oligodendroglial death mediated by glutamate receptors contributes to demyelinating reactions. In the present study, we screened 368 single-nucleotide polymorphisms (SNPs) in 55 genes or gene clusters coding for cytokines, cytokine receptors, suppressors of cytokine signaling (SOCS), complement factors and glutamate receptors for association with MS in a Spanish–Basque resident population. Top-scoring SNPs were found within or nearby the genes coding for SOCS-1 (P ¼ 0.0005), interleukin-28 receptor, alpha chain (P ¼ 0.0008), oncostatin M receptor (P ¼ 0.002) and interleukin-22 receptor, alpha 2 (IL22RA2; P ¼ 0.003). The SOCS1 rs243324 variant was validated as risk factor for MS in a separate cohort of 3919 MS patients and 4003 controls (combined Cochran–Mantel–Haenszel P ¼ 0.00006; odds ratio (OR) ¼ 1.13; 95% confidence interval (CI) ¼ 1.07–1.20). In addition, the T allele of rs243324 was consistently increased in relapsing-remitting/secondary progressive versus primary-progressive MS patients, in each of the six data sets used in this study (PCMH ¼ 0.0096; OR ¼ 1.24; 95% CI 1.05–1.46). The association with SOCS1 appears independent from the chr16MS risk locus CLEC16A. Genes and Immunity (2012) 13, 21–28; doi:10.1038/gene.2011.44; published online 30 June 2011

Keywords: multiple sclerosis; SOCS1; cytokine; genetics; single-nucleotide polymorphism

Introduction been known as susceptibility locus since the early 1970s, it was not until 2007, when the first non-HLA genetic risk Multiple sclerosis (MS) is a chronic (CH) inflammatory factors were unequivocally identified via a genome-wide demyelinating disorder of the central nervous system of association study (GWAS).2 Subsequently, seven more unknown etiology, and represents the most common cause GWAS have led to the identification of around 15 of non-traumatic neurological disability in young adults. validated non-HLA risk loci for MS, including among The observed rates of familial aggregation of MS reflected others IL2RA, IL7R, CD58, EVI5 and CD40.1 In addition, a in the increased risk of siblings, second- and third-degree meta-analysis of GWAS3 identified the additional loci CD6, relatives to develop the disease, as well as twin studies, TNFRSF1A and IRF8, and three further loci with sugges- collectively reject a Mendelian trait as the driving force for tive evidence arising from this study were subsequently susceptibility, but are reconcilable with a polygenic, validated as genuine MS risk factors, that is, IL12A, multifactorial mechanism.1 Although the human leukocyte MPHOSPH9 and RSG1.4 All non-HLA MS-susceptibility antigen (HLA) gene cluster on chromosome 6p21.3 has alleles known so far are relatively common in the population and contribute only modestly to overall risk (odds ratios (OR) of 1.1–1.3). Correspondence: Dr K Vandenbroeck, Neurogenomiks Laboratory, In the present study, we report the results of a haptag Department of Neuroscience, Universidad del Paı´s Vasco (UPV/ screen primarily focusing on cytokine, cytokine receptor EHU), Edificio 205, Planta–1, Parque Tecnolo´gico de Bizkaia, 48170 genes and associated signal transduction factors that also Zamudio (Bizkaia), Spain. E-mail: [email protected] covered a small selection of ionotropic glutamate receptors Received 17 March 2011; revised 5 May 2011; accepted 11 May 2011; and transporters. The latter category of genes was published online 30 June 2011 included based on the observation that glutamate- SOCS1 is associated with MS K Vandenbroeck et al 22 mediated glial injury contributes to white matter pathol- success rate of 97.23%). Risk alleles and their frequencies, ogy as seen in MS.5 The study was performed in two P-values for disease association and ORs of the 20 most phases, with a primary screen of 368 single-nucleotide strongly associated SNPs (Pp0.021), arising from the polymorphism (SNPs) in a Spanish Basque resident case– primary screen, are represented in Table 2. The most control collection of 462 MS patients and 470 controls. Four strongly associated SNPs were found in or nearby the top-scoring SNPs were subsequently analyzed in a genes coding for SOCS1, interleukin-28 receptor, alpha validation cohort composed of five independent sample chain (IL28RA), oncostatin M receptor (OSMR), inter- collections from European and Northern-American origin, leukin-7 receptor (IL-7R) and interleukin-22 receptor, which together included 3919 MS patients and 4003 alpha 2 (IL22RA2) with ORs of around 1.37 and controls. One SNP, rs243324, located in the 50-regulatory uncorrected P-values of 0.003–0.0005. Experimentwise, region of suppressor of cytokine signaling-1 (SOCS1)was none of these associations withstood Bonferroni correc- identified and validated as novel risk factor of MS. tion. Combined single-marker and haplotype analyses highlighted SOCS1 and IL28RA as most strongly associated gene loci (corrected P-values o0.05; Table 3). Of Results seven IL7R tagSNPs typed, rs6897932 emerged as the single most strongly associated one, in confirmation with Details of sample collections used in the present study earlier reports.2,6,7 Detection of this established risk SNP are provided in Table 1. The primary screen of 384 SNPs may attest to the suitability of this case–control collection was performed in the Bilbao collection. A total of 368 to uncover new risk loci with similar or higher OR. SNPs (mean call frequency, 97.25%) was successfully The four new top-scoring SNPs that have not been genotyped for 462 cases and 470 controls (genotyping reported before (IL7R rs6897932 was excluded) were

Table 1 Clinical and demographic features of controls and patients used in the study

Population Controls Cases

Number Male/Female/ Number Male/Female/ RR and SP/PP/ Age-at-onset EDSS nd (%) nd (%) other/nd (%) (average±s.d.) (mean±s.d.)

Bilbao 470 27.9/70.9/1.2 462 28.6/71.4/0.0 81/9.5/2/7.5 30.9±10.3 3.1±2.0 Barcelona 825 46.1/52.8/1.1 708 36.7/63.3/0.0 78/21.2/0.8 30.8±10.2 4.3±2.7 Andalucı´a 1237 30.2/67.5/2.3 1131 29.9/60.7/9.5 80/1/0.9/18.1 29.4±9.9 3.2±1.8 Madrid 832 44.9/53.1/2.0 680 34.5/63.5/2.0 83.2/9.9/0.6/6.3 28.9±8.9 2.9±2.3 UCSF African Americans 658 29.0/69.6/1.4 895 21.0/77.4/1.6 81.7/7/7/4.3 32.8±9.7 4.3±1.6 UCSF Whites 451 33.0/67.0/0.0 505 32.1/67.9/0.0 79.8/3.2/17/0 33.3±9.3 2.04±1.96

Abbreviations: EDSS, expanded disability status scale; nd, not determined, PP, primary progressive; RR, relapsing–remitting; SP, secondary progressive; UCSF, University of California, San Francisco.

Table 2 Twenty most strongly associated SNPs in the primary screen of the Bilbao cohort (Pp0.02)

Gene SNP Chromosome; chromosome Major/minor Risk Cases, risk Controls, risk P-value OR (95% CI) position (location in gene) allelea allele allele % allele %

SOCS1 rs243324 Chr16; 11262471 (50-UTR) C/T T 0.55 0.47 0.0005 1.38 (1.18–1.62) IL28RA rs1416834 Chr1; 24359246 (fourth intron) A/G G 0.41 0.33 0.0008 1.38 (1.17–1.63) OSMR rs3805558 Chr5; 38886646 (first intron) G/A G 0.75 0.69 0.0017 1.38 (1.16–1.65) IL7R rs6897932 Chr5; 35910332 (sixth exon) C/T C 0.76 0.70 0.0022 1.38 (1.15–1.66) IL22RA2 rs202573 Chr6; 137515365 (fifth intron) G/A A 0.33 0.27 0.0026 1.36 (1.14–1.62) IL28RA rs7520329 Chr1; 24371498 (second intron) C/A A 0.30 0.24 0.0026 1.37 (1.14–1.64) GRIK3 rs12067006 Chr1; 37097522 (sixth intron) T/C C 0.29 0.22 0.0034 1.375 (1.14–1.66) IL7R rs6890853 Chr5; 35888068 (50-UTR) G/A G 0.75 0.69 0.0047 1.34 (1.12–1.61) IL31 rs7310689 Chr12; 121217577 (30-UTR) G/A G 0.945 0.91 0.0067 1.66 (1.21–2.28) OSMR rs12657342 Chr5; 38936906 (seventh intron) C/G C 0.79 0.74 0.013 1.31 (1.09–1.59) IL32 rs2015620 Chr16; 3063562 (30-UTR) T/A A 0.34 0.29 0.015 1.27 (1.07–1.51) OSMR rs2278324 Chr5; 38917360 (third intron) C/A A 0.145 0.11 0.015 1.40 (1.10–1.79) IL1RL2 rs17637748 Chr2; 102208147(eighth intron) A/G A 0.57 0.52 0.016 1.25 (1.07–1.47) LIF rs3761427 Chr22; 28974826 (50-UTR) C/G C 0.65 0.60 0.017 1.26 (1.07–1.49) IL1RL2 rs2287040 Chr2; 102219412 (eleventh intron) G/A G 0.68 0.63 0.018 1.26 (1.06–1.49) OSMR rs10040172 Chr5; 38941045 (ninth intron) A/G G 0.17 0.13 0.018 1.36 (1.08–1.70) OSM rs11089441 Chr22; 28984464 (30-UTR) G/T G 0.64 0.58 0.018 1.25 (1.06–1.48) IL15 rs1589241 Chr4; 142784539 (first intron) C/T C 0.68 0.63 0.019 1.26 (1.06–1.49) IL17B rs353268 Chr5; 148737396 (first intron) G/A A 0.09 0.06 0.019 1.52 (1.12–2.06) IL1F6 rs1867828 Chr2; 113485091 (30-UTR) G/A A 0.25 0.20 0.021 1.30 (1.07–1.57)

Abbreviations: CI, confidence interval; OR, odds ratio; SNP, single-nucleotide polymorphism; UTR, untranslated region. aMajor and minor alleles in unaffected controls; reference alleles according to NCBI Reference Assembly.

Genes and Immunity SOCS1 is associated with MS K Vandenbroeck et al 23 Table 3 Ranking of most strongly associated haplotypes and SNPs following permutation testing in the Bilbao seta

Rank Locus SNP or haplotype block Allele or haplotype Case/control P-value 10 000 Permutation frequencies P-value

1 SOCS1 rs193779–rs243324 GT 0.28/0.21 0.0003 0.019 2 SOCS1 rs243324 T 0.55/0.47 0.0005 0.026 3 SOCS1 rs193779–rs243324 GC 0.45/0.53 0.0005 0.030 4 IL28RA rs1416834 G 0.41/0.33 0.0007 0.050 5 IL28RA rs1416834–rs7552086 AA 0.59/0.67 0.0010 0.063 6 IL7R rs6890853–rs1494558–rs3777090– AGATTT 0.235/0.30 0.0011 0.071 rs1494555–rs6897932–rs1053496 7 OSMR rs3805558 G 0.75/0.69 0.0017 0.098 8 OSMR rs3805558–rs834006–rs420444–rs3792856 AAAT 0.24/0.30 0.0017 0.104 9 IL7R rs6897932 C 0.76/0.70 0.0019 0.115 10 IL28RA rs7520329 A 0.30/0.24 0.0025 0.156 11 IL22RA2 rs202573 A 0.33/0.27 0.0027 0.163

Abbreviations: CI, confidence interval; OR, odds ratio; SNP, single-nucleotide polymorphism. aAnalysis includes main single marker hits from Table 2 and additional SNPs found in haplotype blocks with the main hits as determined with the solid spine method (44 included SNPs). 10 000 Permutation test on single markers and haplotype blocks performed with Haploview 4.2.

Table 4 Replication of the four top-scoring SNPs from the Bilbao screen in the validation collections, and combined analysisa

Gene SNP Validation cohort Combinedb

Risk allele PCMH OR (95% CI) PCMH OR (95% CI)

SOCS1 rs243324 T 0.0027 1.10 (1.03–1.18) 0.00006 1.13 (1.07–1.20) IL28RA rs1416834 Ac 0.06 1.06 (1.00–1.14) —d —d OSMR rs3805558 G 0.52 1.02 (0.95––1.10) 0.09 1.06 (0.99–1.13) IL22RA2e rs202573 A 0.42 1.03 (0.96–1.11) 0.06 1.07 (1.00–1.15)

Abbreviations: CI, confidence interval; OR, odds ratio; SNP, single-nucleotide polymorphism. aGiven its well-established association with MS, IL7R rs6897932 was not further validated. bCombined validation and Bilbao data sets. cRisk allele in the validation cohort is opposite to that identified in the Bilbao collection (Table 2). dBreslow–Day test for heterogeneity of effect between the populations was significant (P ¼ 0.0002). eThe African-American data set was excluded because of Hardy–Weinberg disequilibrium in the controls (P ¼ 0.0006). analyzed in the validation cohort comprising five show significant association in any of the individual or independent collections totaling 3919 MS patients and combined validation collections. The A allele of IL22RA2 4003 controls (Table 1). Each population-specific data set rs202573 was significantly associated in the Bilbao was considered as a distinct stratum for testing marker (Table 2) and Andalucı´a(P ¼ 0.04, OR ¼ 1.14 (95% CI association by means of the Cochran–Mantel–Haenszel 1.00–1.29)) data sets, but was not significant in any of the test. Table 4 summarizes the results of the validation and remaining or the combined validation data sets. combined analyses. SOCS1 rs243324 emerged as the only As these SNPs were selected as haplotype-tagging SNP significantly associated in the validation cohort markers, they are unlikely to represent the ultimate

(validation PCMH ¼ 0.0027; combined PCMH ¼ 0.00006). causative variants. As linkage disequilibrium (LD) Similar to the Bilbao data, the rs243324 T allele emerged architecture varies between populations from European as the significantly associated risk allele in the Madrid or African ancestry (http://www.hapmap.org), we also (P ¼ 0.0004; OR ¼ 1.30 (95% confidence interval (CI) analyzed the association patterns of the top-scoring SNPs 1.12–1.51)) and Andalucı´a(P ¼ 0.045; OR ¼ 1.13 (95% CI upon exclusion of the African-American data set. This 1.00–1.27)) data sets, whereas nonsignificant similar increased the strength of the association for SOCS1 trends were found in the remaining three collections rs243324 in both the validation (PCMH ¼ 0.0005; OR ¼ 1.16 (data not shown; Breslow–Day P-value of 0.14 for the (95% CI 1.07–1.26)) and combined (including Bilbao data full-validation cohort). IL28RA rs1416834 displayed a G set) cohorts (PCMH ¼ 0.000004; OR ¼ 1.19 (95% CI 1.11– allele association pattern opposite to that seen in the 1.29)), whereas it did not reinforce association of the Bilbao data set, in both the Barcelona (P ¼ 0.05; OR ¼ 0.86 other SNPs. (95% CI 0.75–1.00)) and Madrid (P ¼ 0.004; OR ¼ 0.79 SOCS1 is located on chromosome 16 at a distance of (95% CI 0.68–0.93)) collections. Owing to this hetero- around 70 kb from the confirmed MS risk gene CLE- geneity effect, the Breslow–Day test for rs141684 in C16A.2,8 We investigated LD between a selection of SNPs combined primary and validation data sets was highly in this gene that recently emerged as good markers for significant (P ¼ 0.0002; Table 4). OSMR rs3805558 did not association with MS;3,8 that is, rs11865121, rs12708716,

Genes and Immunity SOCS1 is associated with MS K Vandenbroeck et al 24 Table 5 Chromosome 16 MS risk loci CLEC16A and SOCS1 are independent

Collection CLEC16A SNPs Position Risk allele; % cases/% controls P-value OR (95% CI) D0/r2 with rs243324

Bilbao rs11865121 11074189 C; 0.69/0.68 0.64 1.04 (0.87–1.24) 0.34/0.054 rs12708716a 11087374 A; 0.64/0.61 0.24 1.11 (0.93–1.31) 0.29/0.052 rs6498169 11156830 G; 0.37/0.35 0.25 1.10 (0.93–1.31) 0.23/0.030 Madrid rs2903692a 11146284 G; 0.66/0.59 0.0002 1.35 (1.15–1.58) 0.26/0.047 rs6498169 11156830 G; 0.39/0.32 0.0007 1.35 (1.13–1.61) 0.25/0.029 Andalucı´a rs2903692a 11146284 G; 0.65/0.60 0.0003 1.26 (1.11–1.42) 0.31/0.063 rs6498169 11156830 G; 0.36/0.34 0.095 1.11 (0.98––1.26) 0.27/0.034

Abbreviations: CI, confidence interval; OR, odds ratio; SNP, single-nucleotide polymorphism. ars12708716 and rs2903692 are proxies (D’ ¼ 1, r2 ¼ 0.96; SNP Annotation and Proxy Search, SNAP, http://www.broadinstitute.org/mpg/ snap/).

its proxy rs2903692 and/or rs6498169, and the most Table 6 SOCS1 rs243324T allele frequencies in RR/SP and PP strongly associated SNP in SOCS1, rs243324. This patients exercise was performed in the Bilbao, Madrid and Andalucıá collections (Table 5). None of the CLEC16A Collection RR/SP PP P-value OR (95% CI) SNPs were significantly associated with MS in the Bilbao cohort, whereas rs2903692 was associated in both the UCSF African-American 0.48 0.45 0.61 1.10 (0.76–1.59) Madrid and Andalucıá collections. Analysis of the LD UCSF whites 0.53 0.50 0.76 1.12 (0.55–2.26) Bilbao 0.56 0.48 0.14 1.41 (0.89–2.21) patterns of each of these CLEC16A SNPs with SOCS1 Barcelona 0.56 0.50 0.09 1.25 (0.96–1.61) rs243324, confirmed that SOCS1 is an independent MS Andalucıá 0.56 0.50 0.58 1.27 (0.55–2.94) 0 2 risk locus (D ¼ 0.25–0.31; r ¼ 0.029–0.063) in each of the Madrid 0.58 0.53 0.29 1.23 (0.84–1.80) three data sets. This is likely to be related to the presence Combined 0.54 0.49 0.0096a 1.24 (1.05–1.46)a of an area showing a recombination rate of up to 8 cM/ Mb separating both the CLEC16A and SOCS1 LD blocks. Abbreviations: CI, confidence interval; OR, odds ratio; PP, primary rs243324 is located in the SOCS1 50-gene-flanking progressive; RR, relapsing–remitting; SP, secondary progressive; region at 5 kb upstream from the transcription initiation UCSF, University of California, San Francisco. site. In a recent study, SOCS1 appeared to be differen- aCochran–Mantel–Haenszel test statistics; Breslow–Day test for tially expressed in immune cell infiltrates in lesions of heterogeneity, P-value ¼ 0.96. relapsing-remitting (RR) compared with chronic (CH) forms of experimental autoimmune encephalomyelitis (EAE).9 We assessed, therefore, whether allelic patterns systematically been investigated in pre-GWAS MS for SOCS1 rs243324 differ between the RR/secondary candidate gene studies. Thus were excluded from the progressive (SP) and primary progressive (PP) forms of present study cytokine gene loci such as IFNG, IL4, MS. Table 6 shows that the frequency of the T allele is on IL1A/IL1B all of which have been analyzed in extenso average 5.2 (±1.9 s.d.)% higher in RR/SP than in PP MS. before.10 In addition, our screen included a selection of This effect was homogenous over the six data sets genes coding for ionotropic glutamate receptors and included in the present study (Breslow–Day P-value of transporters, which are known to regulate brain gluta- 0.96). Even if it did not reach significance in any mate levels implied in the pathological mechanism individual data set because of relative low numbers of driving neuroaxonal cell death.5,11 Glutamate levels in PP patients in each stratum, the effect in the combined the brains of MS patients are at least partially controlled data set was significant (Cochran–Mantel–Haenszel by common genetic polymorphisms, and SNPs in genes P ¼ 0.0096; OR ¼ 1.24 (95% CI 1.05–1.46)). Thus, SOCS1 belonging to the glutamatergic system have been may merit further scrutiny as genetic marker for disease associated with responsiveness to interferon-b therapy course of MS. Analysis of SOCS1 rs243324 in the in MS.12–14 individual and combined gender-stratified study cohorts One SNP, rs243324, located in the 50-regulatory region revealed absence of any significant gender effects (data of SOCS1, emerged from the combined primary screen not shown). and validation study in 44300 MS patients and 44400 controls as a new marker for susceptibility to MS (P ¼ 6.01E-05). SOCS1 functions canonically to impede Discussion Type I and II interferon signaling by binding predominantly to phosphorylated, IFNAR1- or IFNGR-associated Over the last 4 years, GWAS studies have rapidly and JAK proteins through its SH2 domain, thus abrogating successfully outperformed candidate gene studies as STAT activation and dampening ensuing expression of, primary gene-hunting tool for MS risk. The unequivocal among others, the immunoregulatory major histocom- identification via GWAS of susceptibility loci such as patibility complex class I and II, and CD40 genes.15 A IL7R, IL2RA and IL12A has authenticated cytokine and role for SOCS1 in demyelinating conditions has been cytokine receptor genes as a genuine class of MS risk inferred mostly from EAE studies. Upregulation of loci.2,4,6,7 Hence, the present study was designed with as SOCS1 mRNA in EAE is primarily restricted to infiltrat- primary goal the scrutiny through a haptag approach of ing mononuclear cells,16 whereas spinal chord SOCS1 those cytokine and cytokine receptor genes that have not mRNA levels are higher at the peak stage of CH EAE

Genes and Immunity SOCS1 is associated with MS K Vandenbroeck et al 25 (C57BL/6J; myelin oligodendrocyte glycoprotein-in- when both SNPs were considered jointly in these Spanish duced compared with RR-EAE (SJL/J; bovine myelin- data sets. The LD between both SNPs is moderate induced).17 Transgenic mice overexpressing SOCS1 in (D0 ¼ 0.765; r2 ¼ 0.11). Haplotype analysis revealed that oligodendrocytes develop EAE with accelerated onset the rs243324–rs441349 haplotype TC describes the accompanied by enhanced early inflammation; an effect association with MS marginally better than rs243324 related to decreased responsiveness of oligodendrocytes alone, but much better than rs441349 alone (10 000 to early protective effects of interferon-g.18 Administra- permuted P ¼ 0.0001 (TC haplotype), 0.0004 (rs243324) tion of the SOCS1 mimetic, tyrosine kinase-inhibitor and 0.026 (rs441349)). An extensive fine mapping of the peptide, suppressed development of acute EAE in New- SOCS1 gene area including a cluster of four small genes Zealand White mice, protected SJL/J mice with RR-EAE composed of PRM1-3 and transition protein-2, is against relapse,19 and reduces disease severity in C57BL/ currently being performed. Protamines and transition 6 mice with CH-EAE.9 T-cell-specific SOCS1-deficient protein-2 facilitate condensation of genetic material mice are essentially resistant to EAE because of prefer- within the developing spermatid.24 Though no functional ential differentiation of CD4( þ ) T cells into Th1 rather associations with autoimmunity/inflammation/MS than Th17, the Th type known to be essential for EAE.20 seem to have been reported (PubMed search), the Berard et al.9 analyzed expression of SOCS1 in the CH coordinate expression of this multigene locus24 may and RR forms of EAE, induced in the same mouse strain warrant joint investigation in functional studies center- (C57BL/6) using the same myelin antigen (myelin ing upon SOCS1. oligodendrocyte glycoprotein). In both models, SOCS1 IL22RA2 encodes a soluble, secreted receptor for IL-22. was predominantly expressed in macrophages. The A SNP located at a distance of 8.9 kb at 30 from the proportion of Mac-1 þ macrophages expressing SOCS1 IL22RA2 gene, rs276474, was recently found to be in lesions at the peak stage of RR disease was associated with MS in a combined Swedish and significantly higher than that observed in CH-EAE, and Norwegian cohort.25 In the present study, the IL22RA2 coincided with reduced expression of the macrophage SNP rs202573 located in the fifth intron was associated in effector molecule inducible nitric oxide synthase.9 Col- the individual Bilbao and Andalucı´a collections, but was lectively, this data indicates that increased macrophage not significant in the total cohort. Both SNPs are located SOCS1 expression at the RR peak stage may lead to 17.6 kb apart and occur in linkage equilibrium (r2 ¼ 0.013, disengagement of pro-inflammatory effector pathways, D0 ¼ 0.20; SNAP 1000 Genomes Pilot 1 SNP data set). thus promoting remission in the RR–EAE model. With rs1416834, located in the fourth intron of IL28RA, showed this data in mind, we compared SOCS1 rs243324 allele an allelic pattern of association in the Bilbao collection counts in PP versus RR/SP MS patients. Though the opposite to that seen in the Madrid and Barcelona present study was not a priori designed to assess the collections. Interestingly, a recent GWAS identified impact of genetic variation upon clinical course of MS, rs4649203 in the promoter region of IL28RA as novel the total number of PP (n ¼ 347) versus RR/SP (n ¼ 3545) susceptibility locus for psoriasis.26 rs1416834 and patients available via the six sample collections (Table 1) rs4649203 are separated by a 33-kb interval that contains was sufficiently high to allow 73–81% power for a 15-cM/Mb recombination hotspot encompassing the detection of genetic effects conferred by rs243324 with area around the second exon, and, hence, are not in LD OR ¼ 1.25, under additive or multiplicative models, (r2 ¼ 0.033, D0 ¼ 0.28; SNAP 1000 Genomes Pilot 1 SNP respectively. This analysis showed that the rs243324T data set). Although it is possible that the association of allele is consistently increased in RR/SP versus PP IL28RA rs1416834 with MS represents a spurious finding, patients (Table 6). Thus, even if independent confirma- it cannot be excluded that its allelic ‘flip-flop’ association tion is required to firmly validate these findings, and pattern with MS risk is due to sample or population further functional studies are indispensable to assess the variation in inter-locus correlation with the real causative impact of rs243324 on SOCS1 expression, the SOCS1 variant.27,28 Given the fact that MS and psoriasis share locus appears to emerge from this study as a potential the TYK2-susceptibility locus,23,26,29 fine mapping of the marker for clinical course of MS. Functional polymorph- IL28RA region is indicated to verify whether this locus isms altering transcriptional activity of the SOCS1 gene represents a second shared genetic determinant. have been reported,21,22 and one of these, rs33977706, In conclusion, this study has identified and validated located at position À820, occurs in moderate LD with SOCS1 as risk factor for MS. Together with other rs243324 (r2 ¼ 0.19, D0 ¼ 0.80; http://www.broadinstitute. validated risk factors such as CD6, IRF5 and IRF8,3,30,31 org/mpg/snap/ in 1000 Genomes Pilot 1 SNP data set). it is part of an increasing series of MS-susceptibility A recent study analyzed a number of genes with genes of which the products have established roles in functional relationships to the validated MS locus IL7RA, defense against infectious pathogens and in the innate/ and identified a series of SNPs upstream from the SOCS1 adaptive immune response. The SOCS1 rs243324 asso- locus associated with susceptibility to MS.23 The most ciation follows the principle of what is emerging as the strongly associated SNP arising from the combined data gamut of hitherto identified risk SNPs in multifactorial sets used in that study, rs441349, is located in the conditions, that is, low penetrance, high frequency of the intergenic area between protamine-2 (PRM2) and PRM1, risk allelic variant in unaffected subjects and low OR at a distance of 16.8 kb centromeric from rs243324 and at of 1.1–1.3. To this, the potential existence of numerous 21.8 kb from the SOCS1 transcription initiation site.23 We epistatic, epigenetic and random effects, as well as typed rs441349 in both the Madrid and Andalucı´a etiological heterogeneity should be added.1 The indivi- collections (data not shown), and found that, similar to dual value of the SOCS1 SNP in predictive modeling is, Zuvich et al.,23 the C allele constituted the risk allele thus, likely to be limited, but may contribute to strength- associated with MS (PCMH ¼ 0.01). The effect of rs441349 en aggregate models of multiple genetic variants that was less strong than that of rs243324 (PCMH ¼ 0.0002) estimate well-defined clinical parameters of MS. Finally,

Genes and Immunity SOCS1 is associated with MS K Vandenbroeck et al 26 the identification of SOCS1 as MS risk factor may actuate Illumina screen targeted drug development programs. The availability of A customized Bead Array Matrix SNP panel was the SOCS1 mimetic peptide tyrosine kinase-inhibitor manufactured at Illumina Customer Service (San Diego, peptide with demonstrated beneficial effects in EAE is CA, USA). SNPs were characterized for each sample illustrative for the effectiveness of this approach.9,19 in multiplexes of 384 following the Golden Gate protocol of Illumina. Two hundred fifty nanograms (5 mlat 50 ng mlÀ1) of total genomic DNA were used per genotyping reaction. Bead Arrays Matrixes were re- Patients and methods solved at the Illumina’s BeadStation-500GX system. Patients and controls Intensity files obtained from the BeadStation-500GX Demographic and clinical details of sample collections were decoded into genotyping data with the BeadStudio used in the present study are provided in Table 1. The v 3.0 software (Illumina). original screen was performed in the Bilbao collection including 462 MS patients residing in the Spanish Basque Validation of top-scoring SNPs and genotyping of CLEC16A country (Hospital de Basurto) and 470 healthy controls SNPs provided by the Basque Biobank of the Fundacioń Vasca The top-scoring SNPs were replicated in the validation de Innovacioń e Investigacioń Sanitaria. The validation collections using the following Taqman SNP Geno- cohort consisted of 3919 MS patients and 4003 healthy typing Assays (Applied Biosystems, Carlsbad, CA, control subjects contributed by the following centers; USA): C_1004275_10 (SOCS1, rs243324) C_68971_10 Barcelona (Hospital Vall d’Hebron), Madrid (Hospital (IL28RA, rs1416834), C_27503774_10 (OSMR, rs3805558) Clıńico S Carlos), Andalucıá (Hospital Virgen Macarena, and C_629495_10 (IL22RA2, rs202573). The CLEC16A Sevilla; Hospital Carlos Haya, Ma´laga; Hospital Clıńico, SNPs rs2903692 and rs6498169 were genotyped in the Hospital Virgen de las Nieves and Blood Bank, Granada) Madrid and Andalucıá collections using Taqman SNP and University of California San Francisco (African- Genotyping Assays C_15941578_10 and C_29080959_10, American and white collections). All patients were respectively. The Taqman assays were performed accord- ascertained to have definite MS according to the Poser ing to the manufacturer’s instructions. The CLEC16A or McDonald criteria.32,33 Patients and controls were SNPs rs11865121, rs12708716 and rs6498169 were typed included on the basis of informed consent, and the study in the Bilbao collection using the iPLEX Sequenom was approved by local Ethics Committees. MassARRAY platform of the Spanish National Genotyp- ing Center (CEGEN, Santiago de Compostela, http:// www.cegen.org). All genotyping success rates were Selection of SNPs 495%. The concordance between genotyping procedures Haplotype-tagging SNPs were determined using the (Golden Gate versus TaqMan) was tested on a subset of Multimarker Tagger algorithm implemented in HapMap samples and amounted to 98.9%. on the CEU cohort (r2 cutoff 0.8; minimum allele frequency 0.2; HapMap Release #19). Wherever possible, Data management and statistical analysis non-synonymous SNPs were force-included in the Raw data were analyzed for comparison of allele and haptag selection (34 non-synonymous SNPs). The follow- genotype counts using PLINK version 1.05.34 The ing genes or gene clusters were included in the analysis: Cochran–Mantel–Haenszel test implemented in PLINK IL22RA1–IL28RA (chr 1; 9 SNPs); GRIK3 (chr 1; 13 NPs); was used to calculate an average OR arising from the IL23R–IL12RB2 (chr 1; 22 SNPs); IL19–IL20–IL24 (chr 1; 5 combination of distinct case–control data sets. Hetero- SNPs); IL1RL2–IL1RL1–IL18R1–IL18RAP (chr 2; 16 SNPs); geneity of ORs was assessed through the Breslow–Day IL1F7–IL1F9–IL1F6–IL1F8–IL1F5–lL1F10–IL1RN (Chr 2; test, also implemented in PLINK. A w2-test was used to 30 SNPs); IL17RE–IL17RC (Chr 3; 4 SNPs); IL17RB (Chr ascertain that genotype distributions fulfilled the criteria 3; 4 SNPs); IL12A (Chr 3; 3 SNPs); IL21 (Chr 4; 3 SNPs); of Hardy–Weinberg equilibrium. Haplotype analysis IL15 (Chr 4; 4 SNPs); GRIA2 (Chr 4; 3 SNPs); IL7R (Chr 5; was performed with Haploview 4.2 (ref. 35). Unless 7 SNPs); SLC1A3 (Chr 5; 2 SNPs); LIFR (Chr 5; 4 SNPs); otherwise indicated, P-values are uncorrected. Power OSMR (Chr 5; 14 SNPs); C9 (Chr 5; 5 SNPs); C6 (Chr 5; 7 was calculated using the CATS power calculator at SNPs); IL31RA–IL6ST (Chr 5; 13 SNPs); IL13 (Chr 5; 2 http://www.sph.umich.edu/csg/abecasis/CaTS/.36 Link- SNPs); IL9 (Chr 5; 2 SNPs); IL17B (Chr 5; 2 SNPs); GRIA1 age disequilibrium patterns between SNPs were ana- (Chr 5; 22 SNPs); IL17A–IL17F (Chr 6; 11 SNPs); IL20RA lyzed with the SNP Annotation and Proxy Search tool at (Chr 6; 10 SNPs); IL22RA2 (Chr 6; 4 SNPs); IL7 (Chr 8; 2 http://www.broadinstitute.org/mpg/snap/.37 SNPs); C5 (Chr 9; 6 SNPs); IL15RA (Chr 10; 6 SNPs); SLC1A2 (Chr 11; 14 SNPs); IL18BP (Chr 11; 2 SNPs); GRIA4 (Chr 11, 11 SNPs); IL18 (Chr 11, 3 SNPs); IL23A Conflict of interest (Chr 12; 2 SNPs); IL22 (Chr 12; 2 SNPs); SOCS2 (Chr 12; 4 SNPs); IL31 (Chr 12; 3 SNPs); IL17D (Chr 13; 5 SNPs); The authors declare no conflict of interest. IL25 (Chr 14; 3 SNPs); IL16 (Chr 15; 8 SNPs); IL32 (Chr 16; 4 SNPs); SOCS1 (Chr 16; 3 SNPs); IL21R (Chr 16; 8 SNPs); IL17C (CHr 16; 2 SNPs); SOCS3 (Chr 17; 2 SNPs); EBI3 Acknowledgements (Chr 19; 2 SNPs); TMED1 (Chr 19; 1 SNP); IL12RB1 (Chr 19; 3 SNPs); IL28B–IL28A–IL29 (Chr 19; 12 SNPs); IL11 We express our gratitude to all patients and control (Chr 19; 4 SNPs); GRIK1 (Chr 21; 13 SNPs), IL17RA (Chr subjects participating in this study. We thank the Basque 22; 6 SNPs); MIF (Chr 22; 8 SNPs); LIF (Chr 22; 3 SNPs); Biobank of the Basque Foundation for Health Innovation OSM (Chr 22; 4 SNPs). and Research for supplying DNA samples. This study

Genes and Immunity SOCS1 is associated with MS K Vandenbroeck et al 27 was supported by grants to K Vandenbroeck from the 14 Comabella M, Craig DW, Morcillo-Sua´rez C, Rıó J, Navarro A, European Community’s Seventh Framework Program Fernańdez M et al. Genome-wide scan of 500 000 single- (FP7/2007-2013) under grant agreement no 212877 nucleotide polymorphisms among responders and nonre- (UEPHA*MS; www.reem.es/uepha-ms/), and from the sponders to interferon beta therapy in multiple sclerosis. Arch Gobierno Vasco (Ref. IT512-10; Convocatoria ‘Grupos de Neurol 2009; 66: 972–978. Investigacioń 2010–2015’); by grants to A Aransay from 15 Baker BJ, Akhtar LN, Benveniste EN. SOCS1 and SOCS3 in the Gobierno Vasco (Ref. MV-2005-1-13; ‘Convocatoria de the control of CNS immunity. Trends Immunol 2009; 30: Programas de Perfeccionamiento y Movilidad del Perso- 392–400. 16 Maier J, Kincaid C, Pagenstecher A, Campbell IL. Regulation nal Investigador 2005’), the Department of Industry, of signal transducer and activator of transcription and Tourism and Trade of the Government of the Autono- suppressor of cytokine-signaling gene expression in the brain mous Community of the Basque Country (Etortek of mice with astrocyte-targeted production of interleukin-12 or Research Programs 2005–2006) and from the Innovation experimental autoimmune encephalomyelitis. Am J Pathol Technology Department of the Bizkaia County; by project 2002; 160: 271–288. grant FIS PI10/1985 to E Urcelay; by grants to A Alcina 17 Stark JL, Cross AH. Differential expression of suppressors of from the Fondos Europeos de Desarrollo Regional cytokine signaling-1 and -3 and related cytokines in central (FEDER), Ministerio de Ciencia e Innovacioń (SAF2009- nervous system during remitting versus non-remitting forms 11491) and Junta de Andalucıá (P07-CVI-02551); and by of experimental autoimmune encephalomyelitis. Int Immunol grant FIS PI/081636 to F Matesanz. 2006; 18: 347–353. 18 Balabanov R, Strand K, Goswami R, McMahon E, Begolka W, Miller SD et al. Interferon-gamma-oligodendrocyte interac- tions in the regulation of experimental autoimmune encephalomyelitis. J Neurosci 2007; 27: 2013–2024. References 19 Mujitaba MG, Flowers LO, Patel CB, Patel RA, Haider MI, Johnson HM. Treatment of mice with the suppressor of 1 Oksenberg JR, Baranzini SE. Multiple sclerosis genetics – is cytokine signaling-1 mimetic peptide, tyrosine kinase inhibi- the glass half full, or half empty? Nat Rev Neurol 2010; 6: tor peptide, prevents development of the acute form of 429–437. experimental allergic encephalomyelitis and induces stable 2 The International Multiple Sclerosis Genetics Consortium. remission in the chronic relapsing/remitting form. J Immunol Risk alleles for multiple sclerosis identified by a genomewide 2005; 175: 5077–5086. study. N Eng J Med 2007; 357: 851–862. 20 Tanaka K, Ichiyama K, Hashimoto M, Yoshida H, 3 De Jager PL, Jia X, Wang J, de Bakker PIW, Ottoboni L, Takimoto T, Takaesu G et al. Loss of suppressor of cytokine Aggarwal NT et al. Meta-analysis of genome scans signaling 1 in helper T cells leads to defective Th17 and replication identify CD6, IRF8 and TNFRSF1A as new differentiation by enhancing antagonistic effects of multiple sclerosis susceptibility loci. Nat Genet 2009; 41: IFN-gamma on STAT3 and Smads. J Immunol 2008; 180: 776–782. 3746–3756. 4 The International Multiple Sclerosis Genetics Consortium. 21 Gylvin T, Ek J, Nolsøe R, Albrechtsen A, Andersen G, IL12A, MPHOSPH9/CDK2AP1 and RSG1 are novel Bergholdt R et al. Functional SOCS1 polymorphisms are multiple sclerosis susceptibility loci. Genes Immun 2010; 11: associated with variation in obesity in whites. Diabetes Obes 397–405. Metab 2009; 11: 196–203. 5 Matute C, Domercq M, Sańchez-Go´mez MV. Glutamate- 22 Harada M, Nakahima K, Hirota T, Shimizu M, Doi S, Fujita K mediated glial injury: mechanisms and clinical importance. et al. Functional polymorphism in the suppressor of cytokine Glia 2006; 53: 212–224. signaling 1 gene associated with adult asthma. Am J Respir Cell 6 Lundmark F, Duvefelt K, Iacobaeus E, Kockum I, Wallstro¨mE, Mol Biol 2007; 36: 491–496. Khademi M et al. Variation in interleukin 7 receptor a chain 23 Zuvich RL, McCauleu JL, Oksenberg JTR, Sawcer SJ, De Jager (IL7R) influences risk of multiple sclerosis. Nat Genet 2007; 39: PL, IMSGC et al. Genetic variation in the IL7RA/IL7 pathway 1108–1113. increases multiple sclerosis susceptibility. Hum Genet 2010; 7 O’Doherty C, Kantarci O, Vandenbroeck K. IL7RA polymorph- 127: 525–535. isms and susceptibility to multiple sclerosis. N Eng J Med 2008; 24 Wykes SM, Nelson JE, Visscher DW, Djakiew D, Krawetz SA. 358: 753–754. Coordinate expression of the PRM1, PRM2, and TNP2 8 The International Multiple Sclerosis Genetics Consortium. The multigene locus in human testis. DNA Cell Biol 1995; 14: expanding genetic overlap between multiple sclerosis and 155–161. type I diabetes. Genes Immun 2009; 10: 11–14. 25 Beyeen AD, Adzemovic MZ, Ockinger J, Stridh P, 9 Berard JL, Kerr BJ, Johnson HM, David S. Differential Becanovic K, Laaksonen H et al. IL-22RA2 associates with expression of SOCS1 in macrophages in relapsing-remitting multiple sclerosis and macrophage effector mechanisms and chronic EAE and its role in disease severity. Glia 2010; 58: in experimental neuroinflammation. J Immunol 2010; 185: 1816–1826. 6883–6890. 10 Vandenbroeck K. Cytokine Gene Polymorphisms in Multifactorial 26 Genetic Analysis of Psoriasis Consortium et al. A genome- Conditions. CRC Press: Boca Raton, 2006. wide association study identifies new psoriasis susceptibility 11 Srinivasan R, Sailasuta N, Hurd R, Nelson S, Pelletier D. loci and an interaction between HLA-C and ERAP1. Nat Genet Evidence of elevated glutamate in multiple sclerosis using 2010; 42: 985–990. magnetic resonance spectroscopy at 3T. Brain 2005; 128: 1016– 27 Lin PI, Vance JM, Pericak-Vance MA, Martin ER. No gene is an 1025. island: the flip-flop phenomenon. Am J Hum Genet 2007; 80: 12 Baranzini SE, Srinivasan R, Khankhanian P, Okuda DT, Nelson 531–538. SJ, Matthews PM et al. Genetic variation influences glutamate 28 Clarke GM, Cardon LR. Aspects of observing and claiming concentrations in brains of patients with multiple sclerosis. allele flips in association studies. Genet Epidemiol 2010; 34: Brain 2010; 133: 2603–2611. 266–274. 13 Byun E, Caillier SJ, Montalban X, Villoslada P, Fernańdez O, 29 Ban M, Goris A, Lorentzen AR, Baker A, Mihalova T, Brassat D et al. Genome-wide pharmacogenomic analysis of Ingram G et al. Replication analysis identifies TYK2 as a the response to interferon beta therapy in multiple sclerosis. multiple sclerosis susceptibility factor. Eur J Hum Genet 2009; Arch Neurol 2008; 65: 337–344. 10: 1309–1313.

Genes and Immunity SOCS1 is associated with MS K Vandenbroeck et al 28 30 Vandenbroeck K, Alloza I, Swaminathan B, Antigu¨ edad A, on the diagnosis of multiple sclerosis. Ann Neurol 2001; 50: Otaegui D, Olascoaga J et al. Validation of IRF5 as multiple 121–127. sclerosis risk gene: putative role in interferon beta therapy 34 Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, and human herpes virus-6 infection. Genes Immun 2011; 12: Bender D et al. PLINK: a tool set for whole-genome association 40–45. and population-based linkage analyses. Am J Hum Genet 2007; 31 Kristjansdottir G, Sandling JK, Bonetti A, Roos IM, Milani L, 81: 559–575. Wang C et al. Interferon regulatory factor 5 (IRF5) gene 35 Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and variants are associated with multiple sclerosis in three distinct visualization of LD and haplotype maps. Bioinformatics 2005; populations. J Med Genet 2008; 45: 362–369. 21: 263–265. 32 Poser CM, Paty DW, Scheinberg L, McDonald WI, Davis FA, 36 Skol AD, Scott LJ, Abecasis GR, Boehnke M. Joint analysis is Ebers GC et al. New diagnostic criteria for multiple sclerosis: more efficient than replication-based analysis for two-stage guidelines for research protocols. Ann Neurol 1983; 13: genome-wide association studies. Nat Genet 2006; 38: 209–213. 227–231. 37 Johnson AD, Handsaker RE, Pulit SL, Nizzari MM, O’Donnell 33 MacDonald WI, Compston A, Edan G, Goodkin D, Hartung CJ, de Bakker PI. SNAP: a web-based tool for identification HP, Lublin FD et al. Recommended diagnostic criteria for and annotation of proxy SNPs using HapMap. Bioinformatics multiple sclerosis: guidelines from the International Panel 2008; 24: 2938–2939.

Genes and Immunity