RGMA and IL21R Show Association with Experimental Inflammation and Multiple Sclerosis

ORIGINAL ARTICLE RGMA and IL21R show association with experimental inflammation and multiple sclerosis

R Nohra1, AD Beyeen1, JP Guo2, M Khademi1, E Sundqvist1, MT Hedreul1, F Sellebjerg3, C Smestad4, AB Oturai3, HF Harbo4,5, E Wallstro¨m1, J Hillert6, L Alfredsson7, I Kockum1, M Jagodic1, J Lorentzen2 and T Olsson1 1Department of Clinical Neuroscience, Neuroimmunology Unit, Karolinska Institutet, Stockholm, Sweden; 2Department of Biochemistry and Biophysics, Medical Inflammation Research, Karolinska Institutet, Stockholm, Sweden; 3Danish Multiple Sclerosis Center Copenhagen University Hospital, Rigshospitalet, Copenhagen, Denmark; 4Department of Neurology, Oslo University Hospital, Ulleva˚l, Oslo, Norway; 5Department of Neurology, Faculty Division Ulleva˚l, Oslo University Hospital, University of Oslo, Oslo, Norway; 6Department of Clinical Neuroscience, Division of Neurology, Karolinska Institutet, Stockholm, Sweden and 7Department of Cardiovascular Epidemiology, Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden

Rat chromosome 1 harbors overlapping quantitative trait loci (QTL) for cytokine production and experimental models of inflammatory diseases. We fine-dissected this region that regulated cytokine production, myelin oligodendrocyte glycoprotein (MOG)-induced experimental autoimmune encephalomyelitis (EAE), anti-MOG antibodies and pristane-induced arthritis (PIA) in advanced intercross lines (AILs). Analysis in the tenth and twelfth generation of AILs resolved the region in two narrow QTL, Eae30 and Eae31. Eae30 showed linkage to MOG-EAE, anti-MOG antibodies and levels of interleukin-6 (IL-6). Eae31 showed linkage to EAE, PIA, anti-MOG antibodies and levels of tumor necrosis factor (TNF) and IL-6. Confidence intervals defined a limited set of potential candidate genes, with the most interesting being RGMA, IL21R and IL4R. We tested the association with multiple sclerosis (MS) in a Nordic case–control material. A single nucleotide polymorphism in RGMA associated with MS in males (odds ratio (OR) ¼ 1.33). Polymorphisms of RGMA also correlated with changes in the expression of interferon-g (IFN-g) and TNF in cerebrospinal fluid of MS patients. In IL21R, there was one positively associated (OR ¼ 1.14) and two protective (OR ¼ 0.87 and 0.68) haplotypes. One of the protective haplotypes correlated to lower IFN-g expression in peripheral blood mononuclear cells of MS patients. We conclude that RGMA and IL21R and their pathways are crucial in MS pathogenesis and warrant further studies as potential biomarkers and therapeutic targets. Genes and Immunity (2010) 11, 279–293; doi:10.1038/gene.2009.111; published online 14 January 2010

Keywords: multiple sclerosis; experimental autoimmune encephalomyelitis; autoimmunity; RGMA; IL21R

Introduction demonstrated for type 1 diabetes and MS genes.12 Therefore, cross-disciplinary genetics may be rewarding. Common inflammatory autoimmune disorders, such as Discovery of additional genes contributing to MS and multiple sclerosis (MS), type 1 diabetes and rheumatoid their disease regulatory mechanisms may allow the arthritis (RA) are complex chronic diseases with poorly development of more selective therapies and biomarkers understood etiologies. We are particularly interested in in MS. MS that is a chronic inflammatory disease of the central There are many obstacles in studying genetic regula- nervous system. Both environmental and genetic factors tion of autoimmune disorders in human cohorts, includ- contribute to its etiology.1 The human leukocyte antigen ing limited possibility of functional studies and an complex is a major genetic regulator of MS,2–4 whereas uncontrolled contribution of environmental factors. non-human leukocyte antigen genes are numerous and Positioning of disease regulating loci can also be have low odds ratios (ORs).5,6 Only recently, with achieved using animal models in rodents mimicking analysis of very large cohorts, non-human leukocyte the human diseases in which both genetic and environ- antigen MS genes are starting to be unambiguously mental factors can be controlled. Numerous quantitative identified.7–14 Another important concept is the sharing trait loci (QTL) have previously been mapped using of risk genes between inflammatory diseases,15 as now crosses between inbred rodent strains with diverse susceptibilities to autoimmune inflammatory diseases.16 Recent progress suggests that this strategy is productive Correspondence: Dr R Nohra, Department of Clinical Neuroscience, in revealing susceptibility genes and functional path- Neuroimmunology Unit, Neuroimmunology Unit, CMM, L8:04, ways shared between experimental models and complex Karolinska University Hospital, Stockholm SE-171 76, Sweden. 17 E-mail: [email protected] human disorders. Received 9 June 2009; revised 27 November 2009; accepted 30 Experimental autoimmune encephalomyelitis (EAE), November 2009; published online 14 January 2010 a model for MS, has defined pathogenic mechanisms RGMA and IL21R R Nohra et al 280 underlying neuroinflammation, and has allowed MOG-induced EAE (MOG-EAE)29 and in 465 (DA Â 18 development of treatments for MS. Experimental PVG.1AV1) rats of AIL-G12 subjected to PIA. autoimmune encephalomyelitis induced with myelin Linkage analysis in EAE confirmed two separate QTL oligodendrocyte glycopreotein (MOG) in rats closely overlapping with the loci controlling TNF and IL-6 mimics clinical and pathological features of human MS.19 production (Figure 1b). The first QTL, hereafter named Furthermore, the cytokine orchestration in MS and EAE Eae30, spans 6 Mb between the markers D1Rat217 and correlate well.20–22 Similarly, various animal models for D1Rat270, and showed significant linkage to all clinical RA have been used, with pristane-induced arthritis (PIA) phenotypes in addition to a linkage to the production of being the model of choice for studies on erosive RA and anti-MOG IgG2b (Supplementary Table S1; Figures 1b acute-phase responses in arthritis.23 It best fulfills the and c). Disease susceptibility and increased anti-MOG criteria for diagnosis of RA24 and is characterized IgG2b levels were conferred by the EAE-susceptible DA by pronounced bone and cartilage erosions, presence alleles. The second QTL, Eae31, covering a region of of serum rheumatoid factors and T-cell infiltrations B10 Mb between D1Rat193 and D1Rat68, was linked to in joints.25 all clinical phenotypes and to anti-MOG IgG1, IgG2b and In this study, we investigate a quantitative trait locus total IgG titers (Supplementary Table S1; Figures 1b and on rat chromosome 1, originally identified in a c). The PVG allele drove more severe disease and higher

(LEW.1AV1 Â PVG.1AV1) F2 cross (Lewis Â Piebald-Viral- levels of anti-MOG IgGs. For both QTL, there were Glaxo), which carries variants of gene(s) regulating levels effects of sex as an interactive covariate for all linked of tumor necrosis factor (TNF), interleukin (IL)-6 and IL- disease and immune sub-phenotypes in a complex 1b.26 Interestingly, the QTL overlaps loci that regulate EAE manner. On analysis of female and male rats separately and PIA.27,28 Defining genes behind this region might in Eae30, female rats displayed significant linkage to all therefore unravel genetically controlled pathways that clinical phenotypes, but not to anti-MOG IgGs, whereas regulate inflammation in general. Here we aimed first to male rats displayed significant linkage to incidence, day fine-map candidate genes responsible for the regulation of of onset and anti-MOG IgGs. For the Eae31 locus, female EAE and PIA in vivo,aswellasforin vitro cytokine rats also displayed significant linkage to all clinical production after stimulation with lipopolysaccharide phenotypes, but not to the IgG response. In males, there (LPS), and secondly to determine whether any of the was no linkage to clinical phenotypes, but instead to the human homologous genes associate with MS or ex vivo anti-MOG IgGs (Supplementary Table S1). cytokine production. We have refined this large80-Mb In an analogous linkage study on PIA, we identified QTL into two narrow loci: Eae30 and Eae31/Pia32 using the Pia32, spanning B2.1 Mb from D1Rat193 to D1Got334

tenth (G10) and twelfth (G12) generation of advanced and overlapping with Eae31 and the QTL of IL-6 and intercross line (AIL) subjected to EAE and PIA, respec- TNF. Pia32 linked to disease incidence, onset and disease tively. Subsequent investigation of candidate genes from severity (Supplementary Table S2; Figure 1d). Collec- Eae30 and Eae31 in a Nordic MS case–control cohort tively, the Eae31/Pia32 defines a narrow locus controlling demonstrated association of RGMA and IL21R with MS. two different organ-specific inflammatory diseases.

Confirmation of linkage data in a congenic strain

On the basis of data from the F2 cross, we developed a Results congenic rat strain, PVG.LEW-D1Rat270-D1Rat68 (here- A locus on rat chromosome 1 resolves into two independent after called PVG.LEW) by selectively breeding a frag- QTL that regulate expression of TNF and IL-6 ment from the EAE-susceptible LEW.1AV1 into a genetic A region on rat chromosome 1 was previously linked background of the major histocompatibility complex-

to LPS-induced TNF responses in an F2 cross between identical, but EAE-resistant, PVG.1AV1. We used AIL to the EAE-resistant PVG.1AV1 and EAE-susceptible LE- predict how the congenic strain should behave. Interest- W.1AV1 strains.26 To confirm and refine this region, ingly, in an interactive two-QTL model, we could peripheral blood from 465 rats of the twelfth generation identify additional influences in the region from an

(G12) of an AIL was stimulated with LPS and screened for interactive QTL, intQTL, with Eae30 at marker D1Rat131 TNF and IL-6 production. The linkage analysis defined (Supplementary Table S3; Figure 2a). The allelic combi- two distinct loci regulating IL-6 production, whereas nation at different QTL present in the congenic strain TNF production only linked to the distal locus should drive a more severe disease and higher levels of (Figure 1a). Higher TNF and IL-6 levels were driven by TNF and IL-6 (Supplementary Table S4; Figure 2b). the susceptible dark Agouti (DA) alleles at the distal Accordingly, in vitro experiments with LPS stimulation QTL, whereas higher IL-6 levels were driven by the on peripheral blood mononuclear cells (PBMCs) from

resistant PVG.1AV1 alleles at the proximal QTL. The G12 naive PVG.LEW congenics and PVG have demonstrated AIL thus confirmed and considerably refined previously that PVG.LEW produced higher levels of TNF compared reported linkage26 and provided evidence for at least two with the parental PVG.1AV1 strain (Table 1; Figure 3a). distinct genes that regulate levels of TNF and IL-6. Furthermore, IL-6 production was elevated in PVG.LEW compared with PVG.1AV1 (Table 2; Figure 3b). Several experiments with different MOG batches were per- Refined TNF and IL-6 QTL also regulate susceptibility to formed, in which the PVG.LEW congenic displayed a experimental encephalomyelitis and arthritis higher incidence and mortality and a more severe We next sought to determine whether these loci control- disease course displayed by a higher maximum and ling cytokine production also regulate experimental cumulative EAE score (Table 3; Figure 3c). Thus, the inflammatory diseases. We performed a linkage analysis congenic strain confirmed the influence of the region on

in 794 (DA Â PVG.1AV1) rats of an AIL-G10 subjected to EAE, TNF and IL-6 in accordance with linkage analysis

Genes and Immunity RGMA and IL21R R Nohra et al 281

Figure 1 Log-likelihood (LOD: logarithm of the odds) plots for the quantitative trait loci (QTL) identified regulating myelin oligodendrocyte glycoprotein-induced experimental autoimmune encephalomyelitis (MOG-EAE), pristane-induced arthritis (PIA), anti-MOG antibodies and cytokine levels in rats of advanced intercross line (AIL)-G10 and AIL-G12. All markers used for this linkage are not depicted in these plots. Full list of used markers can be found in the Materials and methods section. (a) Linkage to tumor necrosis factor (TNF; thick black line) and interleukin-6 (IL-6; thick gray line) production after WB-LPS stimulation in rats of AIL-G12 (data for IL-6 is from the female subset). Significance threshold for respective phenotype is depicted with the corresponding color and line type as the phenotype itself. (b) Linkage analysis on AIL-G10 rats with cumulative EAE score in MOG-EAE in G10 with sex as an interactive covariate (thick black line). Significance threshold is depicted as a dotted black line. (c) Linkage to total IgG (dashed black line), IgG1 (thick black line) and IgG2b (thick gray line) production on day 12 of MOG-EAE in G10 with sex as an interactive covariate. Significance threshold for respective phenotype is depicted with the corresponding color and line type as the phenotype itself. (d) Linkage analysis on cumulative score in PIA-G12 with sex as an interactive covariate (thick black line). Significance threshold is depicted as a dotted black line.

in AIL, associating disease susceptibility and severity AIL-G10 and AIL-G12, the shared regulatory region with elevated levels of TNF and IL-6. between Eae30, the QTL for IgG2b and IL-6, was reduced to less than 2.4 Mb, spanning 127.9–130.3 Mb on rat Definition of candidate genes chromosome 1. This locus is gene sparse, harboring only We hypothesized that the linked phenotypes for each a few genes and transcripts, including Apolipoprotein AI QTL were controlled by the same genetic variation in regulatory protein 1, the repulsive guidance molecule each of the QTL. On the basis of the combined data on RGMA, chromodomain helicase DNA-binding protein 2

Genes and Immunity RGMA and IL21R R Nohra et al 282

Figure 2 (a) Interaction analysis using a two-dimensional scan for a two-quantitative trait loci (QTL) model between Eae30 and the epistatic

intQTL. The matrix (genomic position Â genomic position) depicts incidence in experimental autoimmune encephalomyelitis (EAE)-G10 with the lower triangle representing the interactive LOD (logarithm of the odds) scores and the upper one showing additive LOD values. The red areas indicate higher LOD values resulting from the interaction between D1Mit17 in the upper triangle and D1Rat131 in the lower triangle. The scale to the right of the matrix shows the additive LOD values (the right side of the scale ranging 1–10) and the epistatic LOD values (the left side of the scale ranging 1–5). (b) Effect plot illustrating the influence of the interaction between Eae30 and intQTL on incidence. On a scale between 0 to 1, where 0 means no disease incidence and 1 is for complete penetrance of disease with 100% incidence in a treated group. A PVG allele at D1Mit17 (Eae30) in combination with a dark Agouti (DA) allele at D1Rat131 (intQTL) leads to a higher EAE incidence (460%).

Table 1 TNF ELISA data after LPS stimulation of PBMCs in PVG.LEW congenics and PVG

Experiment Strain TNF (pg ml)a Distribution by sex Mean TNF (pg ml) by sex

Females Males Females Males

1 PVG 655.5±31.8 15 0 655.5±31.8 – PVG.LEW 865.9±69.0** 20 0 865.9±69.0** – 2 PVG 268.8±9.4 35 21 262.0±11.9 280.3±15.4 PVG.LEW 391.9±20.7*** 27 18 503.3±25.2*** 450.7±31.1*** 3 PVG 251.3±18.0 17 17 169.0±12.7 333.7±18.0 PVG.LEW 299.1±19.4* 10 9 264.9±21.7** 337.2±29.1 4 PVG 266.7±14.6 38 26 258.2±13.5 293.7±19.9 PVG.LEW 368.4±16.0*** 18 16 376.2±28.2*** 375.1±26.5**

Abbreviations: ELISA, enzyme-linked immunosorbent assay; LEW, Lewis; LPS, lipopolysaccharide; PBMCs, peripheral blood mononuclear cells; PVG, Piebald-Viral-Glaxo; TNF, tumor necrosis factor. aTNF concentrations represented as mean values±s.e.m. as measured in the supernatants of PBMCs stimulated with LPS for 18 h. Difference between PVG.LEW congenics vs PVG parental strains was tested using Kruskal–Wallis and Mann–Whitney’s ranking tests with *, ** and *** corresponding to Po0.05, Po0.01 and Po0.001, respectively. In the first experiment, no male rats were used.

(Chd2), alpha-2,8-sialyltransferase 8B (ST8SiaII) and interleukin-21 receptor (IL21R), general transcription solute carrier organic anion transporter family member factor 3C polypeptide 1 (Gtf3c1), GSG1-like, Xpo6 and 3A1 (Slco3a1) (Figure 4). serine/threonine protein kinase SBK1 (src homology 3 The 10-Mb region of Eae31 is more gene dense, with domain-binding kinase 1). A more exhaustive list of many genes having immunoregulatory functions. Im- genes in these QTL can be retrieved from Ensembl plementing the same strategy, as used for Eae30,we Genome browser (Ensembl genome browser, release 50, focused on the common region shared among Eae31, July 2008; positions 127.9–131 Mb and 183.3–185.3 Mb). Pia32 and the QTL for IL-6, TNF and the anti-MOG antibodies. We thus defined a combined confidence Association of RGMA, IL-4 receptor alpha and IL-21 receptor interval smaller than 2.1 Mb. This region contains the with MS gene for 60S ribosomal protein L13, JmjC domain- Initially, we performed single nucleotide polymorphism containing protein 5 (Jumonji domain-containing (SNP) genotyping of tagSNPs in RGMA being candidate protein 5), gene of interleukin-4 receptor alpha (IL4R), from Eae30, IL4R and IL21R being candidates from Eae31/

Genes and Immunity RGMA and IL21R R Nohra et al 283 Pia32, in a Swedish case–control study (SWE I) consisting of 1018 MS patients and 1215 controls. The association analysis identified some nominally significant associations in all genes (Supplementary Table S5). We also found certain haplotypes modestly associated (Supple- mentary Figure 1). We therefore pursued genotyping of the markers in these haplotypes in additional cases and controls from Sweden (SWE II), Norway (NOR) and Denmark (DEN), a total of 2353 additional cases and 1770 controls (see Supplementary Table S6). For the RGMA gene, the results of single marker association studies in the different cohorts are shown in Table 4. In view of the gender influence observed in the rat experiments for Eae30, we also stratified the human material for gender. We then observed that the difference in allele frequency is mainly apparent for male and not for female patients. In fact, in the combined material (SWE I, SWE II and DEN), we observed a significant heterogeneity in the association between males and females for the rs34925346 marker with Po0.04. The C allele of this marker was more frequent in male patients (13%) than in controls (9.7%; OR ¼ 1.33 95% confidence interval (CI): 1.04–1.69; Po0.005), whereas no difference in the frequency of this allele was noticed in females (11%; Figure 5). Further, there is a significant interaction between the rs34925346 marker and sex, as judged by estimating relative excess risk due to interaction,30 which was 2.45 (95% CI ¼ 0.44–4.99) in an analysis of all cohorts except NOR. To evaluate the significance of our findings, we have estimated the false positive report probability (FPRP) for this reported association, which is a function of the prior probability of association and the power of the study (see Supplementary Information and see Supplementary Table S7). For prior odds ranging from 0.01 to 0.002, the FPRP is less than 0.06 (see Supplementary Table S8), which is less than the FPRP used for genome-wide significance in genome-wide association (GWA) studies.31 For IL4R, the rs2234897, rs1805011, rs1805015 and rs1801275 markers that were part of the associated IL4R haplotype in the SWE I material and rs12102586, which also showed nominal association in SWE I, were typed in the SWE II material. The rs1805011 marker replaced the rs2234900 marker that did not work in the TaqMan assay, these markers were in complete linkage disequilibrium with each other in the HapMap data, in our material; r2 ¼ 0.86. These tested IL4R markers did not show any Figure 3 (a) Mean levels of tumor necrosis factor (TNF) as significant association in the SWE II material (data not measured in supernatant of peripheral blood mononuclear cells shown). The rs1801275 marker that was part of the (PBMCs) stimulated with lipopolysaccharide (LPS) for 18 h. The associated haplotype in the SWE I cohort, and has been figure represents data from experiment two in Table 1. The TNF analyzed in several published investigations,32–34 was protein levels are significantly higher in the PVG.LEW (Piebald- genotyped in the NOR and SWE II materials and an Viral-Glaxo.Lewis) strain compared with PVG. This is also observed in both females and males when analysis was segregated for the overall association analysis was performed including different sex subsets. (b) Mean levels of interleukin-6 (IL-6) as reports in the literature. There was no association for this measured in supernatant of PBMCs stimulated with LPS for 18 h. IL4R marker, nor was there significant heterogeneity The figure represents data from experiment 1 as shown in 2. IL-6 between the materials (Figure 6). Further, given a levels are significantly higher in the PVG.LEW strain compared previous report that suggested that the IL4R association with PVG. When segregated for sex the effect was only observed in is mainly observed among DR2-positive MS patients,32 females with a weak tendency towards the same effect in males. (c) Clinical experimental autoimmune encephalomyelitis (EAE) we repeated the meta-analysis stratified for DR2 for the course in PVG.LEW congenic strains compared with PVG rats. The SWE I, SWEII and NOR materials, but no significant LEW.1AV1 allele in the PVG.LEW congenic rats confers increased association for rs1801275 was observed (data not shown). susceptibility and severity to myelin oligodendrocyte glycoprotein- For the IL21R gene, the markers of the rs2107357– induced experimental autoimmune encephalomyelitis (MOG-EAE) rs80603688–rs2214537–rs961914–rs12934152 haplotype compared with PVG rats. Data shown are from the second were typed in SWE II, NOR and DEN materials experiment as shown in Table 3. Threshold for significance (Table 5). The T allele of rs8060368 marker was less *** corresponds to Po0.001. frequent among patients than controls in both SWE I

Genes and Immunity RGMA and IL21R R Nohra et al 284 Table 2 IL-6 ELISA data after LPS stimulation of PBMCs in PVG.LEW congenics and PVG

Experiment Strain IL-6 (pg ml)a Distribution by sex Mean IL-6 (pg/ml) by sex

Females Males Females Males

1 PVG 133.6±9.7 17 17 104.8±8.4 162.5±14.7 PVG.LEW 192.5±9.9*** 11 8 187.3±11.3*** 199.5±18.5 2 PVG 171.9±19.6 5 8 116.4±22.1 206.5±21.3 PVG.LEW 142.6±8.7 6 8 133.3±19.1 149.6±5.9 LEW 377.2±28.0 12 0 377.2±28.0 –

Abbreviations: Il, interleukin; LEW, Lewis; LPS, lipopolysaccharide; PBMCs, peripheral blood mononuclear cells; PVG, Piebald-Viral-Glaxo. aIL-6 concentrations represented as mean values±s.em. as measured in the supernatants of PBMCs stimulated with LPS for 18 h. Difference between PVG.LEW congenics vs PVG parental strains was tested using Kruskal–Wallis and Mann–Whitney’s ranking tests with *, ** and *** corresponding to Po0.05, Po0.01 and Po0.001, respectively. In the first experiment, no male rats were used. In experiment 1, significance noticed in the whole group of animals is most probably conferred by females in this experiment, as no statistical significance was noticed in males but only a small tendency towards the same direction as for females. Experiment 2 showed the same tendency of higher IL-6 production in females but not in males. The group of animals used in this experiment was too small for a clear effect to be noticed.

Table 3 Clinical data in congenics

Experiment Strain Number (Males/females) Incidence (Males/females) Cumulative scorea Maximum scorea Mortalitya (%)

1 PVG 16 (8/8) 1/16 (0/1) 1.2±1.2 0.1±0.1 0/16 (0) PVG.LEW 23 (12/11) 16/23 (7/9)*** 12.9±4.1*** 1.3±0.2*** 0/23 (0) 2 PVG 21 (12/9) 5/21 (1/4) 12.3±6.2 0.5±0.3 1/21 (5) PVG.LEW 18 (8/10) 16/18 (6/10)*** 81.3±12.7*** 3.4±0.4*** 8/18 (44)** 3 PVG 35 (19/16) 20 (10/10) 8.4±2.1 1.2±0.2 1/35 (3) PVG.LEW 22 (8/14) 18/22 (7/11)* 25.1±5.3** 2.6±0.4** 10/22 (46)*** 4 PVG 10 (0/10) 6/10 7.9±3.0 2.2±0.6 2/10 (20) PVG.LEW 10 (0/10) 9/10 37.3±7.6** 4.1±0.5** 7/10 (70)* LEW 10 (0/10) 10/10** 80.8±2.8*** 5.0±0.0** 10/10 (100)**

Abbreviations: LEW, Lewis; PVG, Piebald-Viral-Glaxo. For experiments 1 and 2, experimental autoimmune encephalomyelitis (EAE) was induced with a simple immunization with recombinant rat MOG (rMOG) in complete Freund’s adjuvant (MOG/CFA: 150 mg experiment 1; 200 mg experiment 2). In experiments 3 and 4, a primary immunization with MOG/CFA (MOG/CFA: 60 mg in experiment 3; 40 mg in experiment 4) was followed by a boosting with rMOG in incomplete Freund’s adjuvant (MOG/IFA: 150 mg experiment 1; 200 mg experiment 2) at day 20 after the original immunization. Differences in EAE outcome between PVG.LEW congenic, the EAE-susceptible LEW.1AV1 and the EAE-resistant PVG.1AV1 parental strains were monitored on a daily basis for 40 days from. Only females were used in experiment 4. aValues are for the whole group of animals in each experiment, both affected and unaffected, of each strain. Values represented are mean values±s.e.m. Statistical analysis for incidence of EAE and mortality were tested using Fisher’s exact test. Cumulative and maximum scores were analyzed with Kruskal–Wallis and Mann–Whitney’s test. Threshold for significance is depicted as * where *, ** and *** correspond to Po0.05, Po0.01 and Po0.001, respectively.

and SWE II. The rs2107357–rs80603688–rs2214537– Correlation of RGMA and IL21R with cytokine expression rs961914–rs12934152 GCCCT haplotype that was asso- As expression of certain cytokines may represent ciated in SWE I was not associated in the other important immune sub-phenotypes in MS, we deter- materials, whereas the GTCCC haplotype was asso- mined whether there were any correlations between the ciated in SWE II and GCGCT haplotype in the NOR disease-associated genotypes and expression levels of material (Table 6). The rs8060368–rs2214537–rs961914– those studied in the rat, that is, TNF and IL6. In addition, rs12934152 CGCT haplotype was positively associated the T helper 1-associated cytokine interferon-g (IFN-g) (OR ¼ 1.14 95% CI ¼ 1.06–1.23; Po0.0009) with MS in a was included in view of its central roles in neuroin- joint analysis including the SWE I, SWE II, NOR and flammation. We thus tested for a correlation between the DEN materials, whereas the TCCC and TGCT haplo- genotype of RGMA, IL4R and IL21R and expression types with the same markers was negatively associated levels of TNF, IL-6, IFN-g in cerebrospinal fluid (CSF) with MS (OR ¼ 0.87 95% CI 0.80–0.96; Po0.004, and and PBMCs of MS patients and in patients with other OR ¼ 0.68 95% CI 0.51–0.90; Figure 7). Allelic frequency non-inflammatory neurological diseases. for the CGCT haplotype was 55.1% in patients and For RGMA, we found no correlation between alleles 51.9% in controls, and for the TCCC haplotype 18.4% of the disease-associated rs34925346 marker and in patients and 20.4% in controls. For prior odds IFN-g expression in the CSF of MS patients (Supple- ranging from 0.01 to 0.002, the FPRP is less than 0.04 mentary Figure 2). However, two markers in the (see Supplementary Information and see Supplemen- disease-associated region of RGMA correlated to tary Table S9). higher IFN-g expression: the G allele of rs6497019

Genes and Immunity RGMA and IL21R R Nohra et al 285 Discussion

120 We here demonstrate two narrow QTL: Eae30, linked to

RGMa (NP_001100994.1)-128.56 124 MOG-EAE, and Eae31/Pia32, linked to both MOG-EAE Chd2 (NP_001100993.1)-128.61 128 Eae30 and PIA. This suggests central nervous system-specific ST8SiaII-129.02 130 mechanisms for Eae30 and disease-shared mechanisms Slco3a1-129.57 IgG2b IL6 for Eae31/Pia32. Furthermore, effects of Eae30 and Eae31/ 135 Pia32 on disease correlate with regulation of cytokine production, implicating underlying mechanisms that 140 involve differential production of cytokines. We could confirm this in the PVG.LEW congenic strain that has a

intQTL higher incidence, higher mortality rate and more severe 150 disease course together with higher TNF and IL-6 production compared with the parental PVG.1AV1 153 strain. Despite the high resolution obtained by the AIL approach, formal proof for single genes controlling these 160 important phenotypes is still lacking in the rat, and will require further experimentation. However, limited sets of RNO1 genes allow association studies in the human population.

170 PVG.LEW Thus, findings from the experimental models allowed the formation of hypotheses on the underlying disease genes. We here provide evidence for an association of RGMA and IL21R with MS. The influence of these genes 180 IL4Rec.alpha- 184.62 in human MS put them in focus in the rat system also. IL-21 Rec.- 184.62 183 IL-6 IgG Breeding of minimal congenic strains are now focused on Gtf3c1- 184.70 185 Pia32 IgG1 these genes, which will allow in vivo functional studies Xpo6- 185.18 Eae31 IgG2b TNF Sbk1- 185.34 190 and testing of therapeutic strategies in rat models. Eae30 is a newly defined QTL. Of the few genes within 193 Eae30, our interest was focused on member A of the family of repulsive guidance molecule domains (RGMA), 200 because of its role in the nervous system. The RGMA protein is a membrane-bound molecule originally defined as an axonal guidance molecule in the visual Figure 4 A summarizing figure showing rat chromosome 1 system.35 The RGMA protein and its receptor, neogenin, (RNO1) with all quantitative trait loci (QTL) depicted along the have thus mainly been implicated in nervous system chromosome. Shadowed regions symbolize the shared areas development.36–39 The role of RGMA in the nervous between different QTL at respective position. Thick dashed line shows the region covered by the PVG.LEW (Piebald-Viral-Glax- system and the fact that Eae30 showed linkage exclu- o.Lewis) congenic. A list of the most important genes harbored sively to EAE and not PIA, made it a particularly within shared QTL is shown to the right (check appendix I for a interesting candidate for studies of association with MS; complete list of genes and transcripts). however, as discussed below variants of the gene may well differentially regulate immune mechanisms. In addition, RGMA is not only expressed in neural cells, but also in macrophages and/or microglia and infiltrat- and C allele of rs725458 (Po6 Â 10À3 and Po2.0 Â 10À2, ing leukocytes, as demonstrated in a spinal cord injury respectively; data not shown). These expression model.40 Owing to the observed gender influence in correlations were acting in an additive manner and EAE, we also stratified our MS material for gender. were only observed for female MS patients (Supple- A meta-analysis showed an association in males with a mentary Figure 2). single marker within RGMA (OR ¼ 1.33) and an associa- In contrast to the positive impact on IFN-g expression, tion with a CA haplotype (OR ¼ 1.31). Interestingly, SNPs lower IL-6 expression in CSF of MS patients was in the associated region of RGMA are in potential associated with the G allele of rs6497019 (Po0.05), again transcription binding sites. Despite the extensive data this correlation is only significant among female MS on RGMA in neural cells, our data in EAE and MS thus patients (Po0.04, data not shown). suggest that variants of RGMA differentially affect No correlation was found between markers of IL4R immunoregulation, associated with proinflammatory and expression of any of the cytokines mentioned above cytokines and antibody responses. The linkage of the (data not shown). Eae30 region to anti-MOG IgG2b suggests that the same For IL21R, the rs8060368–rs2214537–rs961914–rs12934152 gene variant increasing disease susceptibility, also causes TGCT haplotype that was negatively associated a T helper 1 bias of the immune response.41 It is with MS, displayed a reduced IFN-g expression in interesting that SNPs in the disease-associated region of PBMCs of MS patients (Po0.02; Supplementary Figure RGMA show association even with expression of INFg 3). Carriers of the CCTT haplotype showed increased and IL6 in CSF. However, at this point, the role of this expression of INF-g and reduced expression of TNF observation is unclear given that disease association is (Po0.02 and Po5 Â 10À5, respectively; Supplementary with rs34925346 among male patients and the expression Figure 3). This haplotype was not significantly associated correlation is mainly for rs6497019 among female with MS. patients. Downstream events of RGMA may be consid-

Genes and Immunity RGMA and IL21R R Nohra et al 286 Table 4 Test of association RGMA markers to MS in four different Scandinavian case–control materials

SNP Minor Major Patients Controls p alleleic Male Male p allelic Female Female p allelic Test of allele allele test patients controls test patients controls test hetero- male female geneity P MAF n MAF n MAF n MAF n MAF n MAF n

SWE I rs1881842 G A 13.1 1986 0.7 1999.1 0.025 12.7 314 14.9 872 0.347 13.2 1672 15.9 1474 0.031 0.867 rs34925346 C G 11 1880 0.6 1891 0.254 10.2 294 9.9 842 0.865 11.2 1586 10 1376 0.289 0.727 rs997941 G A 6.4 1966 0.3 1972.4 0.336 6.1 312 6.9 858 0.634 6.4 1654 7.2 1452 0.364 1 rs6497019 A G 45.3 1982 2.2 2027.3 0.489 49.4 312 43 872 0.053 44.5 1670 44.9 1494 0.812 0.070 rs10520720 A G 20 1994 1 2014 0.139 23 322 21.6 878 0.620 19.4 1672 22 1498 0.080 0.196 rs725458 T C 31.9 1972 1.6 2003.9 0.073 34.8 316 34.4 838 0.888 31.4 1656 34.6 1490 0.055 0.295

SWE II rs1881842 G A 16.2 2172 15.2 1968 0.372 15.6 674 14.5 498 0.597 16.5 1498 15.4 1470 0.437 0.960 rs34925346 C G 12.2 2124 11.5 1920 0.503 14.2 664 11 498 0.117 11.3 1460 11.7 1422 0.754 0.137 rs997941 G A 5.7 2186 7.4 1980 0.027 4.5 682 9.1 504 0.002 6.3 1504 6.8 1476 0.513 0.021 rs6497019 A G 44.5 2136 45.5 1928 0.516 40 662 44.9 486 0.103 46.5 1474 45.8 1442 0.677 0.107 rs10520720 A G 22.3 2136 22.1 1908 0.871 21.8 664 22.2 478 0.892 22.6 1472 22.1 1430 0.768 0.787 rs725458 T C 32.9 2112 32.6 1858 0.842 30.2 658 32.9 462 0.346 34 1454 32.4 1396 0.367 0.202

DEN rs1881842 G A 17.4 864 14 1004 0.049 15.8 272 13.1 464 0.317 18.1 592 14.8 540 0.141 0.934 rs34925346 C G 11.7 856 9.3 766 0.115 13.3 270 7.5 358 0.017 10.9 586 10.8 408 0.946 0.067 rs997941 G A 5.7 892 3.8 1034 0.044 7.6 276 3.1 480 0.006 4.9 616 4.3 554 0.662 0.066 rs6497019 A G 44.5 848 46 980 0.504 46.6 266 46.8 440 0.959 43.5 582 45.4 540 0.523 0.727 rs725458 T C 31.9 908 31.2 994 0.725 35.3 272 31.6 446 0.310 30.5 636 30.8 548 0.901 0.379

NOR rs1881842 G A 15.6 1068 15.8 1028 0.939 15.1 298 15.8a 1028a 15.9 768 15.8a 1028a 0.95 NA rs34925346 C G 11.9 1060 11.4 1012 0.711 12.8 298 11.4a 1012a 11.6 760 11.4a 1012a 0.89 NA rs997941 G A 6.7 1086 5.4 1042 0.193 5.9 304 5.4a 1042a 6.9 780 5.4a 1042a 0.17 NA rs6497019 A G 47.7 1068 45.1 1028 0.247 44.4 302 45.1a 1028a 49.0 764 45.1a 1028a 0.11 NA rs725458 T C 34.9 1012 30.4 974 0.033 34.1 284 30.4a 974a 35.1 726 30.4a 974a 0.04 NA

Abbreviations: DEN, Denmark; MAF, mutation analysis facility; NOR, Norway; SNP, single nucleotide polymorphism; SWE, Sweden. aAs the sex of the Norwegian controls was not known all controls have been included in the comparison with both male and female Norwegian patients.

ered mechanistically. The RGMA protein acts through receptor (IL21R). Both in rats and humans, these two the activation of a RhoA/Rho kinase-dependent path- receptors are positioned close together (Ensembl genome way activation of myosin II.42–44 Inhibition of Rho family browser, release 50, July 2008). functions has been shown to ameliorate EAE in rats, Although an initial study of SNPs in IL4R suggested a associated with promotion of myelin repair, inhibition of disease association, additional screening in SWE II, NOR leukocyte infiltration into the central nervous system and and DEN did not confirm any association of this gene, a reduced axonal damage.45–47 The strong expression of neither on a haplotype nor on a single marker level. RhoA in active MS lesions and low expression in chronic Neither did meta-analysis of all Nordic materials and MS lesions suggest that RhoA also has a role in MS.47 We previous reports of IL4R associations with MS33,34,48 can thus hypothesize that variants of RGMA through reveal any association. Nor did we find any correlation differential regulation of RhoA result in different between expression levels of TNF, IL-6 and IFN-g and the immune activation and disease outcome. IL4R genotype. We therefore conclude that there is no In contrast to Eae30, Eae31/Pia32 showed linkage with contribution of IL4R variants to the susceptibility of MS. two organ-specific diseases, EAE and PIA. The Eae31/ Variants of IL21R displayed association with MS Pia32 QTL also displayed linkage to total anti-MOG in a meta-analysis of all four Nordic cohorts with antibody, IgG1, IgG2b, TNF and IL-6 levels. This a susceptible CGCT haplotype (P-value o9 Â 10À4; probably reflects an overall quantitative effect of the OR ¼ 1.14) and two protective haplotypes: TCCC responsible gene(s) on the pathogenic autoimmune (P-value o4 Â 10À3;OR¼ 0.87) and TGCT (P-value response. A remarkable overlap between QTL-affecting o4 Â 10À3;OR¼ 0.68). In addition, the disease-protective clinical phenotypes of both EAE and PIA, along with the TGCT haplotype also showed an association with lower production of pro-inflammatory cytokines and MOG expression of IFN-g in PBMCs of MS patients. As antibodies, indicates the presence of shared genetic potential molecular mechanism for these effects, several factors acting on each disease. In the 2.1-Mb region of the SNPs in the associated region of IL21R can shared among Eae31, Pia32, IL-6, TNF and anti-MOG potentially affect the transcription factor-binding sites. antibody QTL, we concentrated on two particular genes: Interestingly, after submission of this paper, and con- the interleukin-4 receptor alpha (IL4R) and interleukin-21 sistent with the disease gene-sharing theme, as observed

Genes and Immunity RGMA and IL21R R Nohra et al 287

Figure 5 Association of rs34925346 in RGMA with multiple sclerosis (MS) among (a) males and (b) females. Fixed effect Mantel–Haenszel analysis and Woolf’s test for heterogeneity were performed in R using the meta.MH command in the rmeta package, P-values were estimated in Unphased with study group as a covariate.62 The analysis was performed using the frequency of the C allele of rs34925346 in males and females separately. No Figure 6 Association of rs1801275 in IL4R with multiple sclerosis (MS) significant heterogeneity between the study groups was observed. in eight patient cohorts. Fixed effect Mantel–Haenszel analysis and Odds ratio (OR) ¼ 1.33 among males (95% confidence interval (CI): Woolf’s test for heterogeneity were performed in R using the meta.MH 1.04–1.69; Po0.006), among females OR ¼ 1.04 (95% CI ¼ 0.89–1.21). command in the rmeta package. The analysis was performed using The frequency of this haplotype was 9.7% among male controls. As frequency of the G allele for rs1801275 (also called Q576R and Q551R). sex was not available for the Norwegian (NOR) controls, this cohort The data have been collected from the studies by Suppaih et al.,33 was not included in the sex stratified analysis. Quirico-Santos et al.,34 and Hackstein et al.48 No significant heterogeneity between studies was detected. The overall odds ratio (OR) was 0.99 (95% confidence interval (CI): 0.91–1.07). in the rat experiments, an association of polymorphisms in IL21R with systemic lupus erythematosus was recently published.49 The reported association with placed on the growing list of shared disease genes systemic lupus erythematosus is not in the same region potentially relevant for the regulation of inflammation of the IL21R gene as the association with MS, however, and inflammatory diseases in general. Both RGMA only one marker has been tested in the region we report and IL21R need reproducibility studies in larger associated in this investigation. The downstream me- materials of MS patients, as is ongoing in large GWA chanisms for the genetic influence remain to be demon- studies. Characterization of RGMA- and IL21R-driven strated. However, a differential effect on disease, as pathways both experimentally and in human material observed here, is plausible in view of the pleiotropic might help in the development of selective therapies and effects of the IL21–IL21R pathway on numerous immune biomarkers. mechanisms, including effects on both CD4 þ and CD8 þ T cells, T1/T2 bias, IL-17 production, and B cells and 50 antibody production. Thus, our data on EAE/PIA and Materials and methods MS, and those on systemic lupus erythematosus49 strongly implicate IL21R variants in the regulation of Experimental animals inflammatory processes in general. Further functional Dark Agouti and Lewis (LEW.1AV1) rats, originally studies on the congenic strain will help our under- obtained from the Zentralinstitut fu¨ r Versuchstierzucht standing of how IL21R regulates chronic inflammation. (Hannover, Germany), and Piebald-Viral-Glaxo Neither RGMA, IL4R nor IL21R has been identified as (PVG.1AV1) rats from Harlan UK (Blackhon, UK) were the gene that is involved in autoimmune diseases in bred in the animal facility at the Center for Molecular GWA scans.7,11,14,31,51,52 The region that we report Medicine at Karolinska Hospital. Advanced intercross associated in the RGMA region has reasonable coverage line breeding was established starting with two pairs of of markers in these scans, although rs34925346 marker the major histocompatibility complex-identical DA and was not included in any of the scans, nor was PVG.1AV1 female founders, respectively, producing F1 stratification-based sex performed in all GWA scans. rats. Seven couples of F1 rats with both DA and

For IL21R, the poor marker coverage in GWA scans in the PVG.1AV1 founders were then used to produce the F2 region we identified as associated with MS may explain generation. The G3–G12 generations were created by why this gene has not associated with autoimmune randomly mating 50 pairs of rats from the previous diseases in GWA studies. generation avoiding brother–sister mating. Three litters

To the best of our knowledge, this is the first evidence with 794 AIL-G10 animals were included in MOG-EAE of association of RGMA and IL21R with MS, supported experiments and 463 AIL-G12 animals were collected by data from experimental models. Furthermore, our from two litters and included in the PIA experiment. data suggest a new immunomodulatory role of RGMA Breeding of the PVG.LEW-D1Rat270-D1Rat68 con- in neuroinflammatory disease. The IL21R gene can be genic rat strain (here named PVG.LEW congenic) was

Genes and Immunity RGMA and IL21R R Nohra et al 288 generated by selective transfer of an approximately PVG.1AV1 females. After the tenth generation, rats 63-Mb fragment from LEW.1AV1 (donor strain) into heterozygous for the whole region were intercrossed to a background of the major histocompatibility complex- produce homozygotes for this region. The congenic identical PVG.1AV1 (recipient) rat strain. Initially, strain, PVG.LEW-D1Rat270-D1Rat68 (N10F1), was used

(PVG.1AV1 Â LEW.1AV1) F1 rats were backcrossed to in all MOG-EAE and in vitro experiments described.

PVG.1AV1 females to produce the N2 generation. Rats Animals were kept in a pathogen-free and climate-

from the N2 generation were genotyped with 11 markers controlled environment in polystyrene cages containing on chromosome 1 spanning from D1Rat270 to D1Rat68. aspen wood shavings with free access to standard rodent For each backcross, one male rat containing this fragment chow and water with regulated 12-h light/dark cycles. from LEW.1AV1 was selected and crossed to two Rats were continuously tested according to a health- monitoring program at the National Veterinary Institute (Statens Veterina¨rmedicinska Anstalt) in Uppsala, Table 5 Test of association of MS to IL21R in four different Sweden. All animal experiments were approved by the Scandinavian case-control materials local ethics committee of Northern Stockholm.

SNP Minor Major Patients Controls p allelic Induction of EAE and clinical evaluation allele allele test Recombinant rat MOG (rMOG) corresponding to amino MAF n MAF n acids 1–125 of the N-terminus was expressed in Escherichia coli and purified to homogeneity by chelate SWE I chromatography.53 Rats were anesthetized with isoflur- rs2107357 A G 14.4 1866 13 2266 0.211 rs8060368 T C 36 1960 41.3 2332 0.045 ane (Forane; Abbott Laboratories, North Chicago, IL, rs2214537 C G 43.2 1966 45.5 2384 0.44 USA) before immunization. In total, 794 AIL-G10 rats rs961914 T C 15.7 1798 17.6 2174 0.203 were immunized subcutaneously in the dorsal tail base rs12934152 C T 21.4 1952 24.3 2336 0.113 with a 200-ml inoculum containing 20 mg rMOG in saline, emulsified 1:1 with incomplete Freund’s adjuvant SWE II (Sigma-Aldrich, St Louis, MO, USA). Congenic rats used rs2107357 A G 14.6 2100 13.1 1948 0.16 rs8060368 T C 38.7 2174 45.5 1968 0.019 in clinical MOG-EAE experiments were age-matched rs2214537 C G 45.1 2170 52.6 1964 0.02 and were aged between 12 and 20 weeks. Congenic rats rs961914 T C 16.6 2166 17 1956 0.817 and control rats were immunized using the same rs12934152 C T 25.2 2164 29.8 1954 0.025 induction protocol as for AIL rats and complemented with 20 mg mlÀ1 of Mycobacterium tuberculosis, strain 37 DEN RA (Difco Laboratories, Detroit, MI, USA). rs2107357 A G 11.2 954 13.2 1060 0.174 rs8060368 T C 40.7 948 46.9 1050 0.146 Different rMOG concentrations were used for congenic rs2214537 C G 50.5 936 51.3 1056 0.868 immunizations depending on the rMOG batch used for rs961914 T C 16.1 910 15.2 1044 0.686 the particular experiment. Rats in experiments 1 and 2 rs12934152 C T 25.7 884 29.1 1034 0.275 (Table 3) were immunized with the same rMOG batch using recombinant rMOG in complete Freund’s NOR adjuvant, with the concentrations 150 and 200 mg per rs2107357 A G 14 1040 14.2 1024 0.937 rs8060368 T C 37.9 1044 43.9 1030 0.133 rat, respectively. Rats used in experiment 4 were rs2214537 C G 47 1042 49.9 1040 0.524 immunized with a second rMOG batch with 40 mg rMOG rs961914 T C 15.3 1060 17.7 1018 0.259 in complete Freund’s adjuvant per rat. In experiment 1, rs12934152 C T 24.6 1062 26 1022 0.621 we used 60 mg rMOG in complete Freund’s adjuvant per rat of a third rMOG batch. A single immunization was Abbreviations: DEN, Denmark; MAF, mutation analysis facility; performed in experiments 1 and 2 as described above. NOR, Norway; SNP, single nucleotide polymorphism; SWE, Owing to the variation of potency among different MOG Sweden. batches, a boosting of immunization was required for

Table 6 Test of association of IL21R haplotypes to MS in four different Scandinavian case–control materials

Haplotype SWE I SWE II DEN NOR

Frequency Frequency P Frequency Frequency P Frequency Frequency P Frequency Frequency P patients controls patients controls patients controls patients controls

GTCCC 17.1 19.4 0.058 19.4 22.6 0.033 20.2 22.5 0.203 19.3 19.8 0.757 GCCTT 3.1 3.1 0.997 1.6 2 0.403 2 1.7 0.592 2.6 2.8 0.787 GCGTT 10.7 12 0.195 12.5 12.2 0.782 11.7 11.6 0.974 10.7 12.2 0.278 GTCCT 7.3 7.2 0.875 7.4 7.7 0.791 7.6 7.4 0.855 7.5 9 0.21 GCCCT 3 1.7 0.007 2.7 3.1 0.511 3.1 2.3 0.266 2.5 1.7 0.25 GTGCT 2.1 2.6 0.326 1.5 2 0.359 1.7 2.2 0.428 1.3 2.2 0.099 ACGCT 14.3 13 0.204 14.4 13 0.287 11.4 13.1 0.234 13.6 13.9 0.863 GCGCT 42.4 41.2 0.401 40.5 37.4 0.091 42.5 39.3 0.148 42.7 38.4 0.049

Abbreviations: DEN, Denmark; MS, multiple sclerosis; NOR, Norway; SNP, single nucleotide polymorphism; SWE, Sweden. The markers included in the haplotypes are rs2107357, rs8060368, rs2214537, rs961914, rs12934152.

Genes and Immunity RGMA and IL21R R Nohra et al 289

Figure 7 Association of rs8060368–rs2214537–rs961914–rs12934152 haplotypes in IL21R with multiple sclerosis (MS) in four Scandinavian case–control studies. Fixed effect Mantel–Haenszel analysis and Woolf’s test for heterogeneity were performed in R using the meta.MH command in the rmeta package. P-values for overall association is from a joint association test in Unphased v3.0.13 using study cohort as a covariate.62 No significant heterogeneity between the study groups was observed. The haplotype frequencies among controls are given in the figure. (a) CGCT haplotype. P-value for association ¼ 0.0009. Odds ratio (OR): 1.14 (95% confidence interval (CI) ¼ 1.06–1.23). (b) TCCC haplotype. P-value for association ¼ 0.004. OR ¼ 0.87 (95% CI ¼ 0.80–0.96). (c) TGCT haplotype. P-value for association: 0.004. OR ¼ 0.68 (95% CI ¼ 0.51–0.90). Danish cohort was not included as the haplotype frequency is less than 1% in this cohort. disease induction in experiments 3 and 4. Rats were score of arthritis, the sum of score over time attained by boosted on day 20 with the same rMOG dose as in the each tested animal. initial immunization but excluding the M. tuberculosis in the boosting experiment for an incomplete adjuvant. Genotyping of the AILs for MOG-EAE and PIA models Animals were monitored daily for clinical EAE signs Genomic DNA was extracted from tail/ear tips. Primers starting from day 7 after immunization to day 40 for microsatellite markers (simple sequence length following a scoring scale as follows: 0, no clinical signs; polymorphisms) spanning the 79-Mb region mapped in 26 1, tail weakness or paralysis; 2, hind leg paraparesis or F2 were selected from two genomic databases: the Rat hemiparesis; 3, hind leg paralysis or hemiparalysis; 4, Genome Database and Ensembl Genome Browser. The tetraplegy or moribund and 5, death. Rats were killed if following 28 simple sequence length polymorphisms severe balance disturbance, weight loss greater than 20% were used for genotyping AIL both in G10 and G12: compared with the day of immunization and/or severe D1Rat217, D1Got101, D1Rat269, D1Rat200, D1Mit137, disease for more than 1 day were noticed. D1Rat321, D1Rat37, D1Rat270, D1Rat183, D1Rat41, D1Rat209, D1Rat273, D1Rat131, D1Rat158, D1Rat51, Induction of pristane-induced arthritis D1Rat139, D1Rat437, D1Rat155, D1Rat193, D1Rat66, Pristane-induced arthritis experiments were carried out D1Rat65, D1Got334, D1Arb19, D1Mit13, D1Rat287, in G12 generation of AIL. Animals aged 120–150 days D1Rat110, D1Got170 and D1Rat68. Forward primers received one intradermal injection at the dorsal base of fluorescently labeled with VIC, NED, PET (purchased the tail of 200 ml pristane (C19H40; Sigma). The animals from Applied Biosystems, Foster City, CA, USA) reverse were visually examined every second day, and arthritis primers and 6-FAM-labeled forward primers (Proligo, was scored for each paw as follows: 0, no joints affected; Paris, France) were amplified by PCR according to a 1, one type of joint affected (redness and/or swelling); standard protocol. The PCR products were separated 2, two types of joints affected; 3, three types of joints using the capillary electrophoresis sequencer (ABI3730) affected; and 4, entire paw affected. The types of joints and analyzed using the GeneMapper v3.7 software examined were peritarsal, intratarsal and ankle joints. (Applied Biosystems). Scores were added, yielding a total score for all four limbs ranging from 0 to 16. Animals were scored until 72 Linkage analysis and statistical analysis of AIL and congenic days post-injection. Signs of chronic arthritis were experiments evaluated from day 40 post-injection, and every 10 days Linkage analysis with interval mapping of rat chromo- by comparing paw maps depicturing inflamed joints some one was performed on data obtained from the at a given time point. Arthritis was classified as chronic MOG-EAE experiment in AIL-G10 rats and PIA experi- when new sites of inflammation appeared over time, as ment in AIL-G12 rats, respectively, using R/qtl software previously described.54 Other macroscopic phenotypes version 2.5.1.55 Interval-mapping analysis was performed analyzed were: incidence, the cumulated number of by implementation of the Haley–Knott regression model affected rats (scoresX1) over time, divided by all tested to identify QTL of main effect.55 Two-dimensional scans rats; severity, the maximal score (1–16) attained by each with a two-QTL model were used to identify epistatic affected rat; day of onset, the first recorded sign of joint QTL. Finally, a fit multiple-QTL model test was inflammation; maximum arthritis, the maximal score performed, providing information regarding the interac- (0–16) attained by each tested animal; and cumulative tion of both an additive and epistatic character between

Genes and Immunity RGMA and IL21R R Nohra et al 290 two loci by making an association between all pairs of from whole blood stimulations were collected after 18 h markers and inter-marker positions along the investi- and stored at À70 1C. Protein quantifications for TNF and gated region of a chromosome. We constructed a fit IL-6 were performed using ELISA. model to confirm the linkage to Eae30 and Eae31 and the observed interactive intQTL using the following formula Protein quantification of variance in R/qtl software: y BEae30 þ intQTL þ We used commercially available ELISA kits for TNF Eae31 þ Eae30: intQTL. On dropping the effect of each (Eli-pair, Biosite, San Diego, CA, USA) and IL-6 QTL and the interaction at a time, an impact of the (Biosource, Camarillo, CA, USA) protein quantifications. particular QTL is determined. For TNF ELISA, 96-well flat-bottom ELISA plates (Nunc) Permutation analysis could not be used for the were coated overnight at 4 1C with a capture anti-TNF determination of significance levels due to the structure antibody and saturated with phosphate buffered saline/ of the AIL. Therefore, a residual threshold approach, 5% bovine serum albumin. Pre-coated and saturated in which each animal’s value for a phenotype was plates were used to measure rat IL-6 protein levels. subtracted from the mean value of this particular Assays were performed according to manufacturers’ phenotype for the family of origin, was applied. recommendation. Cytokine levels from supernatants of Confidence intervals in linkage studies have previously non-stimulated samples were below detectable levels in been defined by a 1-LOD (logarithm of the odds) drop all ELISA assays. from the peak marker in a quantitative trait locus.56 In our case, we applied a more stringent confidence MS patients and healthy controls for association studies interval of 2-LOD drop from respective peak marker to The first Swedish multiple sclerosis case–control study insure a full coverage of potential candidate genes, as (SWE I) consisted of 1018 MS patients (84% females) and compared to the previously suggested 1.8-LOD drop for 1215 blood donor controls (63% females), all originating intercrosses.57 from Sweden or other Nordic countries. The patients Data for ELISA, expression analysis and differences in fulfilled the McDonald criteria58 for definite multiple cumulative and maximum EAE scores in the EAE sclerosis and were recruited by neurologists at the experiments were analyzed using Kruskal–Wallis and Karolinska University Hospital Huddinge and Solna Mann–Whitney ranking tests (JMP, version 7, SAS sites in Stockholm, Sweden. The patients were between Institute, Cary, NC, USA). The binomial clinical EAE 22 and 91 years of age (median: 53 years) and the controls phenotypes, that is, EAE incidence and mortality, were between 21 and 76 years (median: 47 years). For were analyzed using Fisher’s exact test. The thresholds further details see the study by Roos et al.59 for statistical significance were set as follows: In a follow-up study, three new cohorts were used *P-valuep0.05; **P-valuep0.01; ***P-valuep0.001. from Sweden, Norway and Denmark. The second Swedish case–controls study (SWE II) consisted of 705 Anti-MOG antibody measurement newly diagnosed MS patients and 663 age- and sex- Levels of MOG antibodies, IgG1, IgG2b and total IgG matched controls from Sweden, and additional 588 MS were determined by ELISA in blood sera from rats of cases. The median age at onset for the SWE II cases was

AIL-G10 collected on day 12 post immunization with 31.5 years with a range of 11–64 years of age; median age MOG. The 96-well ELISA plates (Nunc, Roskilde, Den- at sampling was 37 years. About 67% of the SWE II cases mark) were coated with 10 mgmlÀ1 of rMOG (100 ml per and 77% of the controls were female. well) overnight at 4 1C and then washed with phosphate The Norwegian cohort (NOR) of 548 MS patients (72% buffered saline/0.05% Tween 20. A solution of 5% fat- females) and 554 blood donor controls (54% females) had free milk in phosphate buffered saline/0.05% Tween 20 a median age of 54 years (range 22–90 years) and 46.6 was used to block free binding sites in the ELISA plates years for the blood donor controls (range: 34–58 years). for 1 h at room temperature. The diluted blood sera and The Danish cohort (DEN) consisted of 512 MS patients control sera, previously determined from pilot studies as (56% females) between 20 and 80 years of age (median: containing high concentration of respective isotype to be 43 years), and 553 blood donor controls (39% females) used for standard curves, were added after washing and between 20 and 78 years of age (median: 40 years). incubated for 1 h at room temperature. An additional Oral and/or written consent was given by all wash was performed before incubation of rabbit anti-rat individuals involved in the study. All human experi- IgG (1:2000), IgG1 (1:1000) or IgG2b (1:2000) (Nordic, ments were approved by the local ethics committees of Tilburg, The Netherlands) for 1 h at room temperature. Stockholm, Denmark and Norway. Unbound antibodies were washed away and plates incubated with peroxidase-conjugated goat anti-rabbit Haplotype tagging antiserum (Nordic) (1:10000) for 30 min at room Tagging SNPs (tSNPs) were identified using genotypes temperature. A final wash was carried out before available at the HapMap database on Caucasian families visualizing the bound antibodies with 3,30,5,50-tetra- for SNPs covering the investigated genes. We used methylbenzidine (Sigma). Reactions were stopped by TAGGER in Haploview v3.32 for identifying tSNPs with incubation with 1 M HCl for 15 min in darkness and r2X0.8 and allowing aggressive two and three marker optical density was read at 450 nm. tagging. A total of 97 tSNPs were used for the mapping of RGMA (36 tSNPs), IL4R (25 tSNPs) and IL21R (36 Stimulation of peripheral blood mononuclear cells with LPS tSNPs) in SWE I. In addition, nine markers were selected Blood samples from rat-tail tips were collected in lithium in regions in which there were large gaps between the Heparin Microtainer tubes (Falcon; Becton-Dickinson, markers in the HapMap data, and in which there did not Franklin Lakes, NJ, USA). Stimulation with LPS was seem to be strong linkage disequilibrium. Furthermore, then performed as reported previously.26 Supernatants two non-synonymous SNPs or markers in potential

Genes and Immunity RGMA and IL21R R Nohra et al 291 transcription binding sites, as judged by an analysis Preparation of PBMCs and CSF cells using the RAVEN program, were included. For a total of Peripheral blood from patients with MS (n ¼ 370; mean 14 markers our genotype assay did not work or was of age: 39 years; range: 16–77 years; 68.1% females and poor quality (success rate o80% or did not follow 31.9% males; 94% with and 6% without IgG oligoclonal Hardy–Weinberg equilibrium; Po0.01). Thus, 36 markers bands in CSF) and individuals with other neurological in RGMA, 24 markers in IL4R and 32 markers in IL21R diseases (n ¼ 90; mean age: 39 years; 72.2% females and were analyzed in the SWE I material (see Supplementary 27.8% males; all without IgG oligoclonal bands in CSF) Table S1 and Supplementary Figure 1). The follow-up were used for our expression studies. Paired CSF cells studies in SWE II, NOR and DAN were only genotyped from MS (n ¼ 322; mean age: 39 years; range: 16–77 years; with the associated tSNPs from SWE I screening, 71.1% females and 28.9% males; 94% with and 6% either through single marker association or haplotype without IgG oligoclonal bands in CSF) and other associations. The RGMA region was genotyped neurological diseases (n ¼ 74; mean age: 39 years; 68.9% with rs1881842, rs34925346, rs997941, rs6497019, rs105- females and 31.1% males; all without IgG oligoclonal 20720 and rs725458. The IL4R region was mapped with bands in CSF) were also included. All MS patients rs2234897, rs1805011, rs2234900, rs1805015, rs1801275 fulfilled the McDonald criteria. The CSF was collected in and rs12102586, whereas IL21R was mapped with siliconized glass tubes or polypropylene tubes, and rs2107357, 8060368, rs2214537, rs961914 and rs12934152. immediately centrifuged to recover the pellet and stored at À80 1C until use for RNA preparation. The peripheral blood was sampled in sodium citrate-containing cell Genotyping of the human material preparation tubes (Vacutainer CPT, Becton-Dickinson). Genotyping of the SWE I cohort took place at the Peripheral blood mononuclear cells were separated by Mutation Analysis Facility, Karolinska Institutet, on a density gradient centrifugation. Cells from the inter- Sequenom MassARRAY SNP Genotyping Platform that phase were collected and washed twice with phosphate uses matrix-assisted laser desorption/ionization time-of- buffered saline. More than 95% of the cells were viable. flight60 for detection of allelic variants. Extension assays Finally, cells were pelleted, frozen on dry ice and stored were based on the IPLEX technique. Primer sequences at À80 1C until use. All human experiments were are available upon request, for further details regarding approved by the local ethics committee of Northern the genotyping see the study by Roos et al.59 A total of Stockholm. 92 out of 106 initially considered SNPs passed quality control. The quality control requirements were Hardy– Relative quantification of mRNA by real-time Weinberg P40.01 and success rate 480%. The genotyp- quantitative PCR ing was further validated using a set of 14 trio families Total RNA was extracted from cell pellets using PicoPure (42 individuals) with genotype data available through RNA isolation kit (Arcturus Bioscience, Mount View, CA, the HapMap consortium; concordance analysis with the USA) according to the manufacturer’s instructions. Sam- HapMap data (97.7% concordance). The follow-up ples were treated for 15 min with DNase (Qiagen Rnase- studies on SWE II, NOR and DEN were genotyped free DNase set, Hilden, Germany) to eliminate contamina- with TaqMan SNP genotyping assays using a 7900 HT tion of genomic DNA. Preparation of complementary Fast Real-time PCR system (Applied Biosystems), as DNA was carried out with 10 mloftotalRNA,random previously described.61 The rs2234900 assay could not hexamer primers (0.1 mg; Gibco BRL, Life Technologies, be designed for TaqMan and was thus replaced Ta¨by, Sweden) and Superscript Reverse Transcriptase with rs1805011, a nearby SNP in complete linkage (200 U; Gibco BRL). Quantitative analysis of messenger disequilibrium. RNA expression was performed with iQSYBR green Supermix (Bio-Rad Laboratories, Hercules, CA, USA) using the iCycler thermal cycler, iQ5 Real-Time PCR Association analysis and genotype-expression correlations detection system. We used Beacon Designer 6.0 software Differences in allele frequencies between MS patients (PREMIER Biosoft International, Palo Alto, CA, USA) to and controls were tested through single marker associa- design the primers for glyceraldehyde-3-phosphate dehy- tion analysis using the model commands in PLINK v drogenase, TNF, IFN-g and IL-6 (primer sequences and 1.04. Haplotype association analyses in blocks with high PCR protocols can be provided on request). Sequencing of linkage disequilibrium were performed in Haploview the different bands (Cybergene AB, Huddinge, Sweden) 3.32. A joint meta-analysis test of association for all tested confirmed homology with the reported sequences for the MS populations with the study cohort, as a covariate human genes. Relative quantification of messenger RNA in Unphased62 was performed using the software was calculated by the standard curve method using the Unphased v3.0.13. Evidence of association was also Bio-Rad iQ5 Optical System Software Version 2.0 with confirmed in Unphased and the software PLINK v1.04 endogenous glyceraldehyde-3-phosphate dehydrogenase (for the haplotype analysis) resulting in similar results. as a background expression control. The standard curves Differences in expression levels between carriers of were created using five serial dilutions (1:10, 1:102, 1:103, different alleles were assessed in Unphased program 1:104 and 1:105) of either tested amplicons for each target after transformation, if necessary, to achieve normal or complementary DNA from human blood cells stimu- distribution of the expression levels. Association be- lated with ConA. tween haplotypes and expression levels was analyzed in the haplo.stats package in R using an additive model. Estimation of the false positive report probability Correlation between single marker alleles and expression We have estimated the probability that the association levels was also confirmed in GraphPad Prism software between MS and Eae30 and Eae31 is false by estimating Version 5.01 using a non-parametric Mann–Whitney test. the FPRP as suggested by Wacholder et al.63 The FPRP

Genes and Immunity RGMA and IL21R R Nohra et al 292 depends on the prior probability of association, the new multiple sclerosis susceptibility loci on chromosomes 12 power of the study and the significance threshold: and 20. Nat Genet 2009; 41: 824–828. 12 International Multiple Sclerosis Genetics Consortium FPRP ¼ 1=ð1 þ posterior odds for true associationÞ (IMSGC). The expanding genetic overlap between multiple sclerosis and type I diabetes. Genes Immun 2009; 10: 11–14. Posterior odds for true association 13 Rubio JP, Stankovich J, Field J, Tubridy N, Marriott M, ¼ prior oddsÂpower=significance threshold Chapman C et al. Replication of KIAA0350, IL2RA, RPL5 and CD58 as multiple sclerosis susceptibility genes in The estimation of these measures is given in the Australians. Genes Immun 2008; 9: 624–630. Supplementary Information. 14 De Jager PL, Jia X, Wang J, de Bakker PI, Ottoboni L, Aggarwal NT et al. Meta-analysis of genome scans and replication identify CD6, IRF8 and TNFRSF1A as new multiple sclerosis Conflict of interest susceptibility loci. Nat Genet 2009; 41: 776–782. 15 Vyse TJ, Todd JA. Genetic analysis of autoimmune disease. The authors declare no conflict of interest. Cell 1996; 85: 311–318. 16 Jagodic M, Kornek B, Weissert R, Lassmann H, Olsson T, Dahlman I. Congenic mapping confirms a locus on rat chromosome 10 conferring strong protection against myelin Acknowledgements oligodendrocyte glycoprotein-induced experimental autoim- 53 This study was supported by grants from the Swedish mune encephalomyelitis. Immunogenetics 2001; : 410–415. 17 Aitman TJ, Critser JK, Cuppen E, Dominiczak A, Fernandez- Research council; the Swedish foundation for neuro- Suarez XM, Flint J et al. Progress and prospects in rat genetics: logically handicapped (NHR), Bibbi and Nils Jensens a community view. Nat Genet 2008; 40: 516–522. Foundation, Montel Williams Foundation, So¨derbergs 18 Steinman L. Assessment of animal models for MS and Foundation, the fp 6 EU programs Neuropromise demyelinating disease in the design of rational therapy. (LSHM-CT-2005-018637 and Euratools (LSHG-CT-2005- Neuron 1999; 24: 511–514. 019015). The Norwegian part of the study was supported 19 Weissert R, Wallstrom E, Storch MK, Stefferl A, Lorentzen J, by Norwegian Foundation for Health and Rehabilitation Lassmann H et al. MHC haplotype-dependent regulation of (2004/2/0125) and Odd Fellow MS society, Norway. The MOG-induced EAE in rats. J Clin Invest 1998; 102: 1265–1273. Norwegian Bone Marrow Donor Registry is thanked for 20 Issazadeh S, Lorentzen JC, Mustafa MI, Hojeberg B, Mussener its collaboration in the establishment of the Norwegian A, Olsson T. Cytokines in relapsing experimental autoimmune encephalomyelitis in DA rats: persistent mRNA expression of control material. proinflammatory cytokines and absent expression of interleukin-10 and transforming growth factor-beta. J Neuro- immunol 1996; 69: 103–115. References 21 Olsson T, Zhi WW, Hojeberg B, Kostulas V, Jiang YP, Anderson G et al. Autoreactive T lymphocytes in multiple sclerosis 1 Ebers GC. Environmental factors and multiple sclerosis. Lancet determined by antigen-induced secretion of interferon-gam- Neurol 2008; 7: 268–277. ma. J Clin Invest 1990; 86: 981–985. 2 Chao MJ, Barnardo MC, Lincoln MR, Ramagopalan SV, 22 Olsson T. Cytokine-producing cells in experimental auto- Herrera BM, Dyment DA et al. HLA class I alleles tag HLA- immune encephalomyelitis and multiple sclerosis. Neurology DRB1*1501 haplotypes for differential risk in multiple 1995; 45: S11–S15. sclerosis susceptibility. Proc Natl Acad Sci U S A 2008; 105: 23 Olofsson P, Nordquist N, Vingsbo-Lundberg C, Larsson A, 13069–13074. Falkenberg C, Pettersson U et al. Genetic links between the 3 Lincoln MR, Montpetit A, Cader MZ, Saarela J, Dyment DA, acute-phase response and arthritis development in rats. Tiislar M et al. A predominant role for the HLA class II region Arthritis Rheum 2002; 46: 259–268. in the association of the MHC region with multiple sclerosis. 24 Arnett FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF, Nat Genet 2005; 37: 1108–1112. Cooper NS et al. The American Rheumatism Association 1987 4 Olerup O, Hillert J. HLA class II-associated genetic suscept- revised criteria for the classification of rheumatoid arthritis. ibility in multiple sclerosis: a critical evaluation. Tissue Arthritis Rheum 1988; 31: 315–324. Antigens 1991; 38: 1–15. 25 Vingsbo-Lundberg C, Nordquist N, Olofsson P, Sundvall M, 5 Olsson T, Hillert J. The genetics of multiple sclerosis and its Saxne T, Pettersson U et al. Genetic control of arthritis onset, experimental models. Curr Opin Neurol 2008; 21: 255–260. severity and chronicity in a model for rheumatoid arthritis in 6 Sawcer S. The complex genetics of multiple sclerosis: pitfalls rats. Nat Genet 1998; 20: 401–404. and prospects. Brain 2008; 131(Part 12): 3118–3131. 26 Xu H, Wallstrom E, Becanovic K, Dahlman I, Lorentzen JC. 7 Hafler DA, Compston A, Sawcer S, Lander ES, Daly MJ, De Identification of rat quantitative trait loci that regulate Jager PL et al. Risk alleles for multiple sclerosis identified by a LPS-induced pro-inflammatory cytokine responses. Scand genomewide study. N Engl J Med 2007; 357: 851–862. J Immunol 2002; 56: 248–253. 8 Lundmark F, Duvefelt K, Iacobaeus E, Kockum I, Wallstrom E, 27 Bergsteinsdottir K, Yang HT, Pettersson U, Holmdahl R. Khademi M et al. Variation in interleukin 7 receptor alpha Evidence for common autoimmune disease genes controlling chain (IL7R) influences risk of multiple sclerosis. Nat Genet onset, severity, and chronicity based on experimental models 2007; 39: 1108–1113. for multiple sclerosis and rheumatoid arthritis. J Immunol 9 Weber F, Fontaine B, Cournu-Rebeix I, Kroner A, Knop M, 2000; 164: 1564–1568. Lutz S et al. IL2RA and IL7RA genes confer susceptibility for 28 Lu S, Nordquist N, Holmberg J, Olofsson P, Pettersson U, multiple sclerosis in two independent European populations. Holmdahl R. Both common and unique susceptibility genes in Genes Immun 2008; 9: 259–263. different rat strains with pristane-induced arthritis. Eur J Hum 10 International Multiple Sclerosis Genetics Consortium Genet 2002; 10: 475–483. (IMSGC). Refining genetic associations in multiple sclerosis. 29 Jagodic M, Becanovic K, Sheng JR, Wu X, Backdahl L, Lancet Neurol 2008; 7: 567–569. Lorentzen JC et al. An advanced intercross line resolves Eae18 11 Australia New Zealand Multiple Sclerosis Genetics Consor- into two narrow quantitative trait loci syntenic to multiple tium (ANZgene). Genome-wide association study identifies sclerosis candidate loci. J Immunol 2004; 173: 1366–1373.

Genes and Immunity RGMA and IL21R R Nohra et al 293 30 Zou GY. On the estimation of additive interaction by use of the 47 Zhang Z, Schittenhelm J, Meyermann R, Schluesener HJ. four-by-two table and beyond. Am J Epidemiol 2008; 168: 212–224. Lesional accumulation of RhoA(+) cells in brains of experi- 31 Wellcome Trust Case Control Consortium. Genome-wide mental autoimmune encephalomyelitis and multiple sclerosis. association study of 14,000 cases of seven common diseases Neuropathol Appl Neurobiol 2008; 34: 231–240. and 3,000 shared controls. Nature 2007; 447: 661–678. 48 Hackstein H, Bitsch A, Bohnert A, Hofmann H, Weber F, Ohly 32 Mirel DB, Barcellos LF, Wang J, Hauser SL, Oksenberg JR, A et al. Analysis of interleukin-4 receptor alpha chain variants Erlich HA. Analysis of IL4R haplotypes in predisposition to in multiple sclerosis. J Neuroimmunol 2001; 113: 240–248. multiple sclerosis. Genes Immun 2004; 5: 138–141. 49 Webb R, Merrill JT, Kelly JA, Sestak A, Kaufman KM, 33 Suppiah V, Goris A, Alloza I, Heggarty S, Dubois B, Carton H Langefeld CD et al. A polymorphism within IL21R confers et al. Polymorphisms in the interleukin-4 and IL-4 receptor risk for systemic lupus erythematosus. Arthritis Rheum 2009; genes and multiple sclerosis: a study in Spanish-Basque, 60: 2402–2407. Northern Irish and Belgian populations. Int J Immunogenet 50 Spolski R, Leonard WJ. Interleukin-21: Basic biology and 2005; 32: 383–388. implications for cancer and autoimmunity. Annual Review of 34 Quirico-Santos T, Suppiah V, Heggarty S, Caetano R, Alves- Immunology 2008; 26: 57–79. Leon S, Vandenbroeck K. Study of polymorphisms in the 51 Plenge RM, Seielstad M, Padyukov L, Lee AT, Remmers EF, interleukin-4 and IL-4 receptor genes in a population of Ding B et al. TRAF1-C5 as a risk locus for rheumatoid Brazilian patients with multiple sclerosis. Arq Neuropsiquiatr arthritis—a genomewide study. N Engl J Med 2007; 357: 2007; 65: 15–19. 1199–1209. 35 Monnier PP, Sierra A, Macchi P, Deitinghoff L, Andersen JS, 52 Baranzini SE, Wang J, Gibson RA, Galwey N, Naegelin Y, Mann M et al. RGM is a repulsive guidance molecule for Barkhof F et al. Genome-wide association analysis of suscept- retinal axons. Nature 2002; 419: 392–395. ibility and clinical phenotype in multiple sclerosis. Hum Mol 36 Matsunaga E, Chedotal A. Repulsive guidance molecule/ Genet 2009; 18: 767–778. neogenin: a novel ligand-receptor system playing multiple 53 Amor S, Groome N, Linington C, Morris MM, Dornmair K, roles in neural development. Dev Growth Differ 2004; 46: Gardinier MV et al. Identification of epitopes of myelin 481–486. oligodendrocyte glycoprotein for the induction of experimen- 37 Matsunaga E, Nakamura H, Chedotal A. Repulsive guidance tal allergic encephalomyelitis in SJL and Biozzi AB/H mice. molecule plays multiple roles in neuronal differentiation and J Immunol 1994; 153: 4349–4356. axon guidance. J Neurosci 2006; 26: 6082–6088. 54 Carlson BC, Jansson AM, Larsson A, Bucht A, Lorentzen JC. 38 Yamashita T, Mueller BK, Hata K. Neogenin and repulsive The endogenous adjuvant squalene can induce a chronic guidance molecule signaling in the central nervous system. T-cell-mediated arthritis in rats. Am J Pathol 2000; 156: Curr Opin Neurobiol 2007; 17: 29–34. 2057–2065. 39 De Vries M, Cooper HM. Emerging roles for neogenin and 55 Broman KW, Wu H, Sen S, Churchill GA. R/qtl: QTL mapping its ligands in CNS development. J Neurochem 2008; 106: in experimental crosses. Bioinformatics 2003; 19: 889–890. 1483–1492. 56 Lander ES, Botstein D. Mapping mendelian factors underlying 40 Schwab JM, Monnier PP, Schluesener HJ, Conrad S, Beschor- Genetics ner R, Chen L et al. Central nervous system injury-induced quantitative traits using RFLP linkage maps. 1989; repulsive guidance molecule expression in the adult human 121: 185–199. brain. Arch Neurol 2005; 62: 1561–1568. 57 Manichaikul A, Dupuis J, Sen S, Broman KW. Poor perfor- 41 Gracie JA, Bradley JA. Interleukin-12 induces interferon- mance of bootstrap confidence intervals for the location of a gamma-dependent switching of IgG alloantibody subclass. quantitative trait locus. Genetics 2006; 174: 481–489. Eur J Immunol 1996; 26: 1217–1221. 58 McDonald WI, Compston A, Edan G, Goodkin D, 42 Hata K, Fujitani M, Yasuda Y, Doya H, Saito T, Yamagishi S Hartung HP, Lublin FD et al. Recommended diagnostic criteria et al. RGMa inhibition promotes axonal growth and recovery for multiple sclerosis: guidelines from the International Panel after spinal cord injury. J Cell Biol 2006; 173: 47–58. on the diagnosis of multiple sclerosis. Ann Neurol 2001; 50: 43 Conrad S, Genth H, Hofmann F, Just I, Skutella T. Neogenin- 121–127. RGMa signaling at the growth cone is bone morphogenetic 59 Roos IM, Kockum I, Hillert J. The interleukin 23 receptor gene protein-independent and involves RhoA, ROCK, and PKC. in multiple sclerosis: a case-control study. J Neuroimmunol J Biol Chem 2007; 282: 16423–16433. 2008; 194: 173–180. 44 Kubo T, Endo M, Hata K, Taniguchi J, Kitajo K, Tomura S et al. 60 Isler JA, Vesterqvist OE, Burczynski ME. Analytical validation Myosin IIA is required for neurite outgrowth inhibition of genotyping assays in the biomarker laboratory. Pharmaco- produced by repulsive guidance molecule. J Neurochem 2008; genomics 2007; 8: 353–368. 105: 113–126. 61 Ekelund E, Saaf A, Tengvall-Linder M, Melen E, Link J, Barker 45 Paintlia AS, Paintlia MK, Singh AK, Singh I. Inhibition of rho J et al. Elevated expression and genetic association links the family functions by lovastatin promotes myelin repair in SOCS3 gene to atopic dermatitis. Am J Hum Genet 2006; 78: ameliorating experimental autoimmune encephalomyelitis. 1060–1065. Mol Pharmacol 2008; 73: 1381–1393. 62 Dudbridge F. Likelihood-based association analysis for nucle- 46 Walters CE, Pryce G, Hankey DJ, Sebti SM, Hamilton AD, ar families and unrelated subjects with missing genotype data. Baker D et al. Inhibition of Rho GTPases with protein Hum Hered 2008; 66: 87–98. prenyltransferase inhibitors prevents leukocyte recruitment 63 Wacholder S, Chanock S, Garcia-Closas M, El Ghormli L, to the central nervous system and attenuates clinical signs of Rothman N. Assessing the probability that a positive report is disease in an animal model of multiple sclerosis. J Immunol false: an approach for molecular epidemiology studies. J Natl 2002; 168: 4087–4094. Cancer Inst 2004; 96: 434–442.

Supplementary Information accompanies the paper on Genes and Immunity website (http://www.nature.com/gene)

Genes and Immunity