Fine-Mapping Resolves Eae23 into Two QTLs and Implicates ZEB1 as a Candidate Gene Regulating Experimental Neuroinflammation in Rat

Background To elucidate mechanisms involved in multiple sclerosis (MS), we studied genetic regulation of experimental autoimmune encephalomyelitis (EAE) in rats, assuming a conservation of pathogenic pathways. In this study, we focused on Eae23, originally identified to regulate EAE in a (LEW.1AV1xPVG.1AV1)F2 cross. Our aim was to determine whether one or more genes within the 67 Mb region regulate EAE and to define candidate risk genes. Methodology/Principal Findings We used high resolution quantitative trait loci (QTL) analysis in the 10th generation (G10) of an advanced intercross line (AIL) to resolve Eae23 into two QTLs that independently regulate EAE, namely Eae23a and Eae23b. We established a congenic strain to validate the effect of this region on disease. PVG alleles in Eae23 resulted in significant protection from EAE and attenuated CNS inflammation/demyelination. Disease amelioration was accompanied with increased levels of Foxp3+ cells in the CNS of the congenic strain compared to DA. We then focused on candidate gene investigation in Eae23b, a 9 Mb region linked to all clinical phenotypes. Affymetrix exon arrays were used to study expression of the genes in Eae23b in the parental strains, where none showed differential expression. However, we found lower expression of exon 4 of ZEB1, which is specific for splice-variant Zfhep1. ZEB1 is an interleukin 2 (IL2) repressor involved in T cell development. The splice-specific variance prompted us to next analyze the expression of ZEB1 and its two splice variants, Zfhep1 and Zfhep2, in both lymph node and spleen. We demonstrated that ZEB1 splice-variants are differentially expressed; severity of EAE and higher IL2 levels were associated with down-regulation of Zfhep1 and up-regulation of Zfhep2. Conclusions/Significance We speculate that the balance between splice-variants of ZEB1 could influence the regulation of EAE. Further functional studies of ZEB1 and the splice-variants may unravel novel pathways contributing to MS pathogenesis and inflammation in general.

The ongoing challenge is to elucidate the molecular mechanisms conferred by the risk genes. Therefore, analyses in a simplified system are advocated.
The genetic dissection of a rat model for MS represents such an approach, with the assumption that there are conserved mechanisms regulating neuroinflammation among species. We study the genetic regulation of myelin oligodendrocyte glycoprotein (MOG) induced experimental autoimmune encephalomyelitis (EAE) in rats due to a close mimicry with MS, with a protracted relapsing disease course, inflammatory infiltrates, prominent demyelination and axonal damage [16]. Consistent with MS, the MHC locus is suggested to be the strongest susceptibility locus in EAE [17]. Furthermore, non-MHC loci that contribute to EAE susceptibility have been identified in several F2 crosses using different strain combinations [18][19][20]. However, the determination of disease-regulating genes requires further fine-mapping and validation approaches.
Here we use an approach combining high-resolution mapping in an advanced intercross line (AIL) with congenic validation and gene expression studies. This approach has successfully been used to identify EAE-regulating genes. An influence of Eae18 on MOG-EAE was confirmed in a congenic strain [21] and the region was fine-mapped in the AIL to identify a cluster of chemokine genes as candidate genes [22]. Further investigation demonstrated association of CCL1 (ENSG00000108702), CCL2 (ENSG00000108691) and CCL12 (ENSRNOG00000029768) with MS [23]. Here we focused on another region, on rat chromosome 17 (RNO17) that was originally identified in a (LEW.1AV1xPVG.1AV1)F2 cross, where linkage analysis revealed a quantitative trait locus (QTL) regulating clinical and histopathological signs of EAE (named Eae23 in this study) [20].
Our aim was to determine whether one or more genes within the 67 Mb region of Eae23 regulate EAE and to fine-map those genes. Eae23 was resolved into two QTLs that independently regulate EAE, namely Eae23a and Eae23b. The effect of Eae23 was confirmed in a congenic strain, DA.PVG-Eae23, and histopathological investigation of the CNS showed that inflammation and demyelination was attenuated compared to DA. We identified ZEB1 (Zfhep, TCF8, dEF1, ZFHX1A, ENSRNOG00000017863) as the most likely candidate gene in Eae23b. ZEB1 is foremost known as a repressor of interleukin 2 (IL2, ENSRNOG00000017348), a cytokine involved in proliferation, apoptosis, cell differentiation and T cell development at multiple stages [24,25]. We further postulate that the balance between splice variants of ZEB1 could influence the regulation of EAE.

Determination of Eae23a and Eae23b using an Advanced Intercross Line
To narrow the confidence interval of Eae23 [20], and confirm the locus in a new strain combination, we investigated the region in the (DAxPVG.1AV1)G10 AIL [26][27][28][29]. In total, 224 out of 772 AIL animals displayed clinical EAE, with an incidence of 29% ( Figure S1). G10 AIL rats were genotyped over 67.7 Mb of RNO17 (13.9-81.6 Mb) using 20 polymorphic microsatellite markers (Table S1). Linkage analysis was performed in females due to a lack of power to detect the QTL in males ( Figure S2). Analysis of clinical phenotypes identified two QTLs, denominated Eae23a and Eae23b (Figure 1).
Eae23a was linked to all clinical phenotypes and constituted a 21 Mb region around peak marker D17Got45 at 47.3 Mb ( Table 1). The confidence interval was likely overestimated owing in part to all 11 microsatellite markers tested between D17Rat12 (33.3 Mb) and D17Got49 (42.7 Mb) being non-polymorphic in the strains used. The allelic effect showed transgressive segregation with alleles from the resistant PVG strain promoting disease in a dominant manner ( Figure 1B). Rats homozygous for DA alleles were protected against EAE with lower incidence (p,0.0002), later onset (p,0.0005), shorter duration of disease (p,0.0004) and with lower maximum and cumulative EAE scores (p,0.0008 and p,0.0006, respectively) when compared to rats with heterozygote or PVG homozygote genotype in Eae23a.
Eae23b was also linked to all clinical phenotypes, with a confidence interval of 9 Mb (57-66 Mb) around peak markers D17Got120 and D17Got70 at 61.8-62.3 Mb (Table 1). Analysis of allelic effects revealed significant heterosis whereby heterozygous alleles differ from either parental allele ( Figure 1C). Rats heterozygous in Eae23b were predisposed to EAE with higher incidence (p,0.0001), earlier onset (p,0.0001), longer duration (p,0.0001) and more severe disease (MAX p,0.0001, CUM p,0.0001) when compared to rats with homozygous DA or PVG genotype in Eae23b. A striking 43% of Eae23b-heterozygous rats were affected by EAE, compared to only 27% of the homozygous rats.
The individual effects of Eae23a and Eae23b on EAE phenotype variation were further supported by a fit multiple QTL model [30] ( Table 1). To estimate the probability of these QTLs contributing to EAE, simulated pedigrees were mapped using a single-QTL model (to force the QTLs to compete for effect). Consistent with the linkage analysis, Eae23a and Eae23b were both probable to influence susceptibility and severity phenotypes ( Table 1).

Confirmation of Eae23 Effects on EAE using Congenic Rats
To confirm the influence of Eae23 on disease we established DA.PVG-Eae23, a congenic strain that carries PVG alleles in Eae23a and Eae23b introgressed on a DA background ( Figure 2). Commonly, the parental DA strain displays a relapsing-remitting disease course with an average onset of EAE around two weeks after immunization, while the parental PVG.1AV1 strain is resistant to the same induction protocol. DA.PVG-Eae23 littermate controls showed no clinical differences to inbred DA (data not shown) and DA were therefore used as control in the experiments. To follow up the AIL findings we first investigated the clinical effects in females during MOG-EAE. In total, 20 of the 25 female rats included in the experiment showed signs of EAE (80% incidence). Although there was no significant difference in susceptibility (INC and ONS) between strains, DA.PVG-Eae23 rats had significantly less severe clinical symptoms compared to DA during both the acute and chronic phase of disease ( Figure 3A). DA.PVG-Eae23 female rats had less severe disease, with lower maximum (p,0.02) and cumulative (p,0.007) EAE scores and shorter duration of disease (p,0.01) ( Table 2). In total, 34 of 43 male rats included in the experiment developed EAE (79% incidence) which was comparable with the incidence in females. DA.PVG-Eae23 males had less severe disease during the acute and chronic phases compared to DA, and had a lower maximum EAE score (p,0.04) ( Figure 3B, Table 2).
To investigate if the heterosis effect of Eae23b detected in the AIL could be captured in the congenic strain, we repeated the experiment with heterozygous DA.PVG-Eae23 rats. Based on the AIL results we expected heterozygous DA.PVG-Eae23 rats to have the full effect of both Eae23a and Eae23b (PVG dominance and heterosis, respectively) with a more severe disease, while homozygous rats should show effect of Eae23a only (PVG dominance) and have an intermediate phenotype. However, there were no significant differences between heterozygous DA.PVG-Eae23 and DA ( Figure 3C). Although the homozygous DA.PVG-Eae23 rats were significantly protected compared to DA (consistent with the previous experiments), more rats in total were affected by EAE (93% incidence) and the congenic rats showed slightly increased disease severity. Since both female experiments were immunized with the same protocol, this slight variation may reflect inter-experiment fluctuation or a seasonal influence (winter compared to spring) on EAE [31]. To understand the lack of effect in heterozygous congenic rats, we selected individuals from the AIL cohort that matched the genotype combinations captured in the congenic rats (Table 3). This analysis was consistent with the results from congenic rats, since the allele combination captured in the heterozygous congenic rats also produced an intermediate phenotype in the AIL that was not significantly different from the phenotypes of rats with DA genotype. We were therefore unable to further investigate the heterosis effect in the congenic strain.

Influence of Eae23 on CNS pathology in Congenic Rats
To evaluate the effects of Eae23 on EAE pathology in the target organ, we investigated inflammation and demyelination in the CNS of DA and DA.PVG-Eae23 female rats at 35 days post immunization (p.i.). Immunohistopathological analysis of the brain and the spinal cord revealed differences in inflammation/ demyelination, as well as in cell activation and infiltration between strains ( Figure 4). The most frequent distribution pattern (41% overall) was characterized by the combined appearance of demyelinated lesions in the optic nerves and the spinal cord. The control group (DA) developed more severe disease compared to the congenic strain (Table 4). However, when we compared rats with the same MAX and ONS, congenic rats showed generally lower susceptibility for CNS inflammation with a lower degree of demyelination and with a lower number of eosinophils and infiltrating CD8 + macrophages [32]. We also observed a lower recruitment of the CD3 + T cell population to the spinal cord in congenic rats (indicated by arrow heads in Figure 4). The distribution pattern of CD3 + T cells, CD8 + macrophages and eosinophils corresponded to the demyelinated lesions in the brain and in the spinal cord. There was significantly less demyelination in both brain and spinal cord of DA.PVG-Eae23 rats compared to DA (p,0.01 and p,0.03, respectively). Interestingly, Foxp3 + T cells were present in higher numbers in the congenic rats compared to DA (indicated by arrow heads in Figure 4).

Influence of Eae23 on regulatory T cell response to MOG recall
To investigate the possibility of an enhanced population of Foxp3 + regulatory T cells (Tregs) in DA.PVG-Eae23, we determined the proportion of Foxp3-expressing T cells in lymph nodes collected 7 days p.i. ( Figure 5). Using flow cytometry, a higher proportion of CD4 + CD25 + Foxp3 + cells was detected in ex vivo lymph node cells from the susceptible DA strain compared to DA.PVG-Eae23 (N = 5 per group; p = 0.008; data not shown). However, in response to MOG restimulation, DA.PVG-Eae23 had a significantly higher up-regulation of the CD4 + CD25 + Foxp3 + population than DA (2.6-and 1.8-fold change, respectively; p = 0.008; Figure 5C).

Expression Studies Identify ZEB1 as a Candidate EAEregulating gene
Eae23a contained 140 genes while Eae23b presented a limited number of candidate genes to investigate further. On the assumption that a shared gene is responsible for linkage between Eae23b and all clinical phenotypes, we restricted the candidate gene list to the region shared among clinical phenotypes. The 5 Mb region (59-64 Mb) contained 30 genes, of which 21 had described functions ( Figure 2). mRNA expression of all candidate genes was investigated in lymph nodes collected day 7 p.i. from parental DA and PVG.1AV1 rats (N = 4 per group) using exon arrays [33]. There were no differences in gene-level expression of the candidate genes between strains that reached genome-wide significance. When we investigated exon-level expression of the genes, there were no differences between parental strains. However, exon 4 of ZEB1 had much lower expression compared to all other exons of this gene, indicating alternative splicing ( Figure 6A). Interestingly, ZEB1 produces two splice variants, Zfhep1 (NP_037296.1, EN-SRNOT00000046013) and Zfhep2 (ZEB1, ENSRNOT000000 24336), for which exon 4 (NP_037296.1 exon 1) is specific for the Zfhep1 variant ( Figure 6B). The resulting protein differs in the first zinc finger cluster ( Figure 6C). This prompted us to investigate if splice-variant specific differences of ZEB1 could be involved in regulation of EAE.
To explore potential differences in ZEB1 using the more sensitive quantitative real-time polymerase chain reaction (RT-PCR) method, we analyzed the mRNA expression of ZEB1 and its target IL2 in lymph nodes collected at day 7 p.i. from MOGimmunized DA and PVG.1AV1 rats (N = 6 per group). There were no significant differences between the strains in either the shared region of ZEB1 or its downstream target IL2 ( Figure 7). However, mRNA expression of the splice variant Zfhep2 was higher in DA compared to PVG.1AV1 (p,0.005) while there was no significant difference in Zfhep1.
To investigate if Eae23 regulates this expression we then measured ZEB1, splice variants and IL2 mRNA expression in lymph nodes collected at day 7 p.i. from MOG-immunized DA and DA.PVG-Eae23 female rats (N = 10 and N = 11, respectively) ( Figure 8A-D).
No significant differences between the strains were seen in ZEB1, Zfhep1 or IL2. However, mRNA expression of the splice variant Zfhep2 was higher in DA.PVG-Eae23 (p,0.018) compared to DA ( Figure 8D), confirming regulation by the Eae23 locus. Similarly, we observed a trend of lower secreted IL-2 in DA.PVG-Eae23 upon recall, although differences were not significant ( Figure S3).
To further investigate the expressional differences in another tissue at a later timepoint, we collected spleens from DA and DA.PVG-Eae23 female rats on day 35 p.i (N = 9 and N = 13) ( Figure 8E-H). The shared region of ZEB1, Zfhep2 and IL2 were significantly upregulated in DA compared to DA.PVG-Eae23 (p,0.001, p,0.002 and p,0.002, respectively), while Zfhep1 was significantly downregulated in DA compared to DA.PVG-Eae23 (p,0.04).
Our data suggests that the balance in expression between Zfhep1 and Zfhep2 could contribute to disease protection in the congenic strain, both in early initiation and ongoing disease.

Discussion
In this study, we used the AIL to resolve Eae23 into two narrow QTLs that independently regulate EAE, namely Eae23a and Eae23b. Linkage of Eae23 in 428 (DA 6PVG.1AV1)G10 AIL female rats. The significance thresholds are indicated in parenthesis and are the maximum LOD scores obtained from the analysis of residual values from family mean data for each phenotype. a Peak marker indicates the microsatellite marker closest to the highest LOD score obtained within the QTL. The fit multiple QTL model was used to validate the independent effects of Eae23a and Eae23b (phenotype = Eae23a + Eae23b + e) and the locations used corresponds to the peak markers. A significant difference in the phenotypic variance defines an independent QTL, and is indicated by * = p,0.05 and ** = p,0.01. b A drop in LOD score of 1.5 from the maximum LOD (LOD support interval) was used to define the confidence intervals. Positions are given in Mb. An influence of Eae23 on neuroinflammation was then confirmed with a congenic strain. These approaches complemented each other to provide evidence for Eae23's importance for neuroinflammation. The congenic strain carrying PVG alleles in Eae23 demonstrated protection from clinical symptoms of disease and attenuated CNS inflammation and demyelination. This was somewhat surprising, as DA alleles in Eae23 were disease protective in the AIL. Further, the combination of Eae23a DA/PVG alleles together with Eae23b DA/PVG alleles (disease-driving in the AIL) produced an intermediate phenotype, suggesting that the regulatory effects exerted by Eae23a and Eae23b may be influenced by other regions in the genome. EAE is a multifactorial disease and several regulatory regions and epistatic interactions that influence EAE have previously been demonstrated in the AIL [27,28]. The AIL contains a genetic heterogeneous background which allows Eae23 to interact with and be influenced by other loci. Conversely, these effects were fixed in the congenic strain in order to study Eae23 in isolation. The AIL is therefore likely to capture more accurate genetic effects in the complex setting and thus enable QTL refinement, while the congenic strain represents the isolated effect of alleles at a single locus and thus provides a good tool for dissecting the molecular mechanisms involved in disease. Out of the 140 genes in Eae23a, butyrophilin subfamily 1 member A1 (Btn1a1) is of particular interest because of its effects in EAE. The extracellular Ig-like domain of butyrophilin shares features with MOG that are suggested to allow molecular mimicry, and butyrophilin can induce or suppress the encephalitogenic T cell response to MOG in DA rats [34]. Additionally, butyrophilin can both prevent and suppress MOG-EAE in mice, with subsequent suppression of IFN-c, IL-2 and IL-12 [35]. The involvement of butyrophilin and other possible candidate genes within Eae23a and their effect on EAE requires further investigation that will be pursued in future studies.
For Eae23b, we defined ZEB1 as a possible candidate gene by using isoform-specific expression profiling of the parental DA and PVG.1AV1 and congenic strains. ZEB1 is a zinc finger homeodomain transcription factor that binds E-box-like sequences at target genes [36] and represses gene transcription by recruiting the C-terminal-binding protein (CtBP) co-repressor [37][38][39]. ZEB1 is foremost known as an interleukin 2 (IL2) repressor [40,41]. Zfhep1, the full-length and fully functional splice-variant of ZEB1, can down-regulate expression of its' own gene in addition to other genes. Conversely, splice-variant Zfhep2 antagonizes this autorepression by competing for the binding sites without repressing ZEB1 expression [42]. Thus, increased expression of Zfhep2 disturbs auto regulation of Zfhep1 leading to subsequent upregulation of Zfhep1 that down-regulates IL2. In line with this, we detected higher expression of Zfhep2 in the lymph node cells of protected DA.PVG-Eae23 congenic rats 7 days after immunization, when the immune response is mounted against autoantigen.
This up-regulation of Zfhep2 in early disease likely led to disrupted Zfhep1 auto-regulation in later disease. Consequently, higher levels of Zfhep1 and lower levels of IL2 were seen in the protected DA.PVG-Eae23 rats 35 days after immunization. However, repression of IL2 expression did not reflect ZEB1 levels, but depended on Zfhep1 expression. This finding probably reflects that the full-length variant contains both zinc-finger domains needed to be fully functional [43]. We therefore postulate that the mechanism by which Eae23b regulates EAE is through regulation of alternative splicing of ZEB1 which in turn regulates IL2.
IL2 is involved in T cell development at multiple stages [24,25,[44][45][46]. The alpha-chain of the receptor for IL2 (IL2RA) is a risk gene for MS and other autoimmune diseases [11,13,47,48]. IL2 was therefore our prime candidate mechanism for ZEB1 influence on neuroinflammation, supported by finding in the congenic strain. We found decreased levels of Foxp3 + cells, which are likely to be Tregs [49,50], and higher levels of CD8 + macrophages in the CNS of disease susceptible DA rats compared to protected congenic rats. This difference was present in rats with the same onset of disease and maximum clinical severity, likely reflecting that DA.PVG-Eae23 rats recovered from the first bout of disease to a larger extent than DA rats and had milder disease in the chronic phase of EAE ( Figure 3). The improved recovery rate observed in congenic rats may reflect the enhanced Treg population. In early disease, we observed a higher potential to expand the Treg population in response to rMOG restimulation in lymph node cells. Consistent with this, a higher number of Tregs were present in the target organ later in disease. Previous studies Table 2. DA.PVG-Eae23 is protected from EAE compared to DA.   have reported EAE-suppressive effects of Tregs in the CNS [51,52]. The higher infiltration rate of Foxp3 + cells in protected rats supports the protective role of Tregs and indicates that downregulated autoimmunity diminished the ''inviting signal'' to recruit peripheral immune cells that cause disease exacerbation. Accordingly, susceptible rats had higher recruitment of macrophages in the CNS and developed more severe inflammation compared to protected rats. This is possibly a consequence of dysregulation of Tregs in DA rats. This speculation is consistent with involvement of IL2. Mice that are deficient of IL2 and IL2RA (CD25) mount strong immune responses that contribute to development of autoimmune disease [53], and reduced IL2 levels correlate with reduced CD4 + CD25 + T cell function [54]. Additionally supporting the involvement of Tregs is that their frequency is normal in MS patients compared to controls, but their suppressive function is diminished [55]. Further study is required to understand the cell populations in the CNS that contribute to disease development and recovery. Therefore, detailed characterization of the kinetics of CNS inflammation in our congenic strain over the course of EAE need to be addressed in future studies. Alternative splicing is a process by which multiple proteins are created from a single pre-mRNA, including exon extensions and deletions [56][57][58]. Bioinformatic analyses suggest that alternative splicing is a significant component in generating diversity of function and predict that the majority of human genes undergo alternative splicing [57]. Splice variants may serve as homeostatic regulators in healthy physiology and disease [59][60][61]. Isoform selection of alternative splice forms may be involved in the pathogenesis of many human diseases and animal models thereof, including cancer and autoimmune diseases [62][63][64][65]. Cytotoxic T lymphocyte-associated antigen 4 (CTLA4, ENSG00000163599) has demonstrated genetic linkage to a number of autoimmune diseases, including MS [66][67][68][69][70], and involves genetic regulation of alternative splice forms [71]. Other genes that regulate disease by differential splice-variant expression are IL2RA and IL7R, which may involve dysregulation of their membrane-bound and soluble forms in MS patients [12,72,73]. For example, the C allele of a risk SNP may contribute to skipping of alternatively spliced exon 6 of IL7RA and alter the balance between splice forms [12]. Taken together, these findings indicate that expression and signaling delivered by specific splice variants may play a critical role in regulating autoimmune disease and illustrates the importance of carefully dissecting gene expression data. Concordingly, a SNP in ZEB1 has recently been suggested to associate with rheumatoid arthritis, another autoimmune disease [74]. We demonstrated that the balance between Zfhep1 and Zfhep2, the splice variants of ZEB1, was altered in EAE-susceptible rats when compared to protected rats. Considering the similarities between MOG-EAE and MS, identifying underlying genes may either reveal genes that regulate MS or identify pathogenic pathways of importance for MS. ZEB1 is a strong candidate because of its regulation of the IL2 pathway, which is important in MS and EAE. Further functional studies of ZEB1 and the splice variants may unravel novel pathways contributing to MS pathogenesis.

Animals
Ethics Statement. All experiments in this study were approved and performed in accordance with the guidelines from the Swedish National Board for Laboratory Animals and the European Community Council Directive (86/609/EEC) under the ethical permit N332/06 entitled 'Genetic regulation, pathogenesis and therapy of EAE, an animal model for multiple sclerosis', which was approved by the North Stockholm Animal Ethics Committee (Stockholms Norra djurförsöksetiska nä mnd). Rats were tested according to a health-monitoring program at the National Veterinary Institute (Statens Veterinä rmedicinska Anstalt, SVA) in Uppsala, Sweden. Inbred DA and PVG.1AV1 rats were originally obtained from the Zentralinstitut für Versuchstierzucht (Hannover, Germany) from which colonies have been established at Karolinska Institutet (DA/Kini and PVG.1AV1/Kini). All animals were bred and kept in 12 h light/ dark-and temperature-regulated rooms. Rats were housed in polystyrene cages containing aspen wood shavings and had free access to standard rodent chow and water.
Advanced Intercross Line. The AIL was established between DA and MHC-identical PVG.1AV1 strains as described previously [28]. Three G10 litters were produced of  which 794 rats were selected for MOG-EAE experiments. The litters were similar in size and sex distribution. Congenic strains. The congenic strains were established from DA and PVG.1AV1 rats using a speed-congenic approach with marker-assisted selection [75]. In brief, DA females were mated with male offspring selected for PVG alleles within Eae23 and against PVG background contamination at 96 microsatellite markers equally spaced throughout the genome (20 centiMorgan intervals). DA females were used throughout the breeding program to ensure that mitochondrial DNA was inherited from DA. A breeding pair selected from the N8 generation was intercrossed to produce homozygous DA.PVG-Eae23 congenic rats, carrying PVG alleles between D17Rat8 at 27.6 Mb and D17Rat37 at 75.8 Mb (Figure 3).

Induction and Clinical Evaluation of EAE
Recombinant rat MOG (amino acids 1-125 from the N terminus) was expressed in Escherichia Coli and purified to homogeneity by chelate chromatography [76]. Rats between 8 and 11 weeks of age were anesthetized with isoflurane (Abbott Laboratories, Abbott Park, IL, USA) and immunized intradermally in the dorsal tail base. Each animal received a 200 ml inoculum containing MOG in phosphate buffered-saline (PBS) (Life Technologies, Paisley, UK) emulsified 1:1 with incomplete Freunds adjuvant (IFA) (Sigma-Aldrich, St. Louis, MO, USA). Each MOG batch had slightly different potency, so titration experiments were performed to determine comparable effective doses. AIL rats were immunized with 20 mg of MOG per rat while congenic rats were immunized with sex-specific MOG doses, 15 and 22 mg/rat in females (depending on MOG batch) and 30 mg/ rat in males. Animals were weighed and clinical signs of EAE were evaluated daily from day 9-35 p.i. as previously described [28]. Phenotypes used for analysis were incidence of EAE (INC), onset of clinical signs (ONS), duration of symptoms (DUR), maximum and cumulative EAE scores (MAX and CUM, respectively). Weight loss (WL) was calculated by subtracting the lowest weight during the experiment from the weight at time of immunization and expressing the difference as a percentage of the initial weight.

Genotype and Statistical Analysis
Genotyping. Genomic DNA was extracted from tail tips according to a standard protocol [77]. Genotypes were determined by polymerase chain reaction (PCR) amplification of microsatellite markers as previously described [78]. For the AIL, forward primers were end labelled with [c-33 P]ATP and the PCR products were size fractionated on 6% polyacrylamide gels and visualized by autoradiography. Primers were obtained from Proligo (Paris, France). For the construction of the congenic strain, fluorophoreconjugated primers were used (Applied Biosystems, Foster City, CA, USA) and PCR products were size fractionated using an electrophoresis capillary sequencer (ABI3730). Genotypes were analyzed using the GeneMapper software (v. 3.7, Applied Biosystems) [79] and all genotypes were manually evaluated by 2 independent observers.
Linkage Analysis of AIL. The genetic map was defined using publicly available genome sequence (http://www.ensembl. org/v.55). Determination of the QTL map position and the confidence interval depends on the correct genomic positions of microsatellite markers. The physical map was used for linkage analysis which enabled comparison of linkage maps of subpopulations. Linkage analysis was performed using R/qtl software version 2.4.1 [80] to identify the QTLs of main effect and twodimensional scans were performed to identify epistatic interactions. For calculation of linkage to phenotypes the binary model was used for INC and Hailey-Knott regression was used for DUR, MAX, CUM and WL, to generate maximum base 10 logarithm of the likelihood ratio (LOD) scores. Hailey-Knott regression was chosen because it could include covariates, and we tested that the model yielded similar results as a non-parametric or two-part model.
To set the significance threshold, we computed the mean for each family and the residual value from the family mean for each rat and repeated the linkage analysis in the residual data set (indicated in parenthesis in applicable tables) [28]. In addition, the maximum LOD scores for the original one-dimensional scans were all higher than 3.4, a suggested cut-off from simulated data sets [30]. The conventional permutation method [81] could not be used to set the significance thresholds due to the difference in family structure of the G10 compared with F2 generations for which it was developed, while the traditional bootstrapping method [82,83] could not be used to determine confidence intervals because of the closely situated QTL. Confidence intervals were defined using the closest marker outside a 1.5 LOD support interval from the QTL peak location [84], approximately equivalent to a 95% confidence interval [82].
Statistical Analysis. Kruskal-Wallis ranking test was used to confirm the influence of peak markers defined in the linkage analysis on phenotypes, and p values # 0.05 after Bonferroni correction [85] were considered significant. The fit multiple QTL model in R/qtl was used to statistically validate the independent effect of each identified QTL. The probability for each QTL was generated by simulating 1000 pedigrees sampled with replacement from the observed individuals (N = 428), and each was mapped using a single-QTL model with Hailey-Knott regression in R/qtl. Probability was calculated as the percentage of times the maximum LOD (above threshold) was located within the confidence interval of each QTL, as previously described [28]. To analyze the experiments with congenic strains, DA.PVG-Eae23 was compared with parental DA. To estimate the influence of allele combinations of Eae23a and Eae23b on disease characteristics, rats from the AIL cohort that carried the genotypes captured in each congenic strain (DA.PVG-Eae23, DA.PVG-Eae23-Heterozygous and DA) were selected. The expected phenotype values for each congenic strain were based on the average phenotype values within the corresponding AIL group. AIL rats with DA genotype were compared to rats with either PVG or heterozygous genotype in Eae23. ONS, MAX, DUR and CUM were analyzed using the Mann-Whitney U-test and Fisher's exact test was used to compare INC. Calculations were carried out using the PC versions of JMP v. 6.0.2 (SAS Institute Inc., Cary, NC, USA). For in vitro experiments data were analyzed using Mann-Whitney's non-parametric ranking test (GraphPad Prism 5.01, GraphPad, San Diego, USA). Median was plotted. Statistical significance was determined as p-value #0.05.

Immunohistochemistry
At day 35 p.i. thirteen DA.PVG-Eea23 rats and nine DA controls were euthanized by asphyxiation using CO 2 and perfused with PBS followed by 4% paraformaldehyde. Immunohistochemical analysis of dissected brains and spinal cords was performed on paraformaldehyde fixed paraffin embedded sections. 0.4 mm thick sections were stained with hematoxylin-eosin and luxol fast blue to detect CNS inflammation and demyelination, respectively. The inflammatory index was determined from the number and size of inflammatory lesions on an average of twenty complete crosssections of the spinal cord and the brain of each animal, as previously described [16].

Expression analysis
Affymetrix. Four DA and four PVG.1AV1 rats were euthanized using CO 2 7 days post-EAE induction, before debut of clinical disease signs. Draining inguinal lymph nodes were collected and samples were prepared as previously described [33]. We used Affymetrix Exon 1.0 ST arrays to assess genome-wide expression. Target labeling, as well as array hybridization, washing and staining were performed as described in the GeneChip Whole Transcript (WT) Sense Target Labeling manual (http://www. affymetrix.com). Arrays were scanned using the GeneChip Scanner 3000 7G (Affymetrix). Probe sequences for all annotated genes within Eae23 are found in Table S2.
The microarray data is available in MIAME-compliant (minimum information about a microarray experiment) format at the ArrayExpress database (http://www.ebi.ac.uk/arrayexpress) [86] under accession code E-MEXP-2237. CEL intensity files were produced using GeneChip Operating Software version 1.4 (Affymetrix) and quality tested using the Affymetrix Expression Console. Probe-level data were normalized using Robust Multiarray Average (RMA) [87]. Detection of differential expression at the exon and gene level was performed in Partek Genomics Suite 6.4 (Partek Inc., St.Louis, MO, USA). Data was summarized at the gene level using a One-Step Tukey's Biweight Algorithm, which reduces the effect of outlier probe-sets (Statistical Algorithms Description Document; Affymetrix-White-Paper). An ANOVA model, using strain, condition and batch as co-factors, was used to generate raw p values, while FDR was used for multiple test corrections [88]. Genes with 5% FDR were classified as differentially expressed.
RNA extraction and RT-PCR. For expression analysis in the parental DA and PVG.1AV1 strains, six females of each strain were immunized with MOG as described above at 12 weeks of age. On day 7 p.i. animals were euthanized using CO 2 and inguinal lymph nodes were dissected out and snap-frozen in micro tubes and stored at 270uC. For congenic strains, 5 DA and 5 DA.PVG-Eae23 female rats were sacrificed using CO 2 at day 7 p.i. and draining inguinal lymph nodes were collected. Lymph nodes were placed in DMEM (Gibco-BRL, Grand Island, NY) enriched with 5% fetal calf serum, 1% L-glutamine, 1% penicillinstreptomycin, 1% pyruvic acid (Life Technologies, Paisley, Scotland) and 50 mM 2-Mercaptoethanol (complete media; Gibco-BRL) before being mechanically separated by passing through a mesh screen with the bolus of a syringe. Cells were spun at 300 g and resuspended in complete media for counting. Cells were allocated to ex vivo flow cytometry, cell culture or RT-PCR. An additional 5 DA and 6 DA.PVG-Eae23 female rats were sacrificed using CO 2 at day 7 p.i. and draining inguinal lymph nodes were collected and snap-frozen in micro tubes and stored at 270uC. Furthermore, 10 DA and 13 DA.PVG-Eae23 females were immunized at 10-12 weeks of age and MOG-EAE was evaluated for 35 days. On day 35 p.i. the rats were euthanized and spleen was dissected out and snap-frozen in micro tubes and stored at 270uC. Tissues were disrupted using Lysing Matrix D tubes (MP Biomedicals, Irvine, CA, USA) on a FastPrep homogenizer (MP Biomedicals).mRNA from cells and homogenates was extracted using RNeasy mini columns (Qiagen Gmbh, Hilden, Germany), including on column DNA-digestion. Reverse transcription was performed using random hexamer primers (Gibco) and Superscript Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA). Real-time PCR was performed using an ABI7900HT Fast Real-Time PCR System (Applied Biosystems Inc., Foster City, CA, USA) with a two-step PCR protocol (50uC for 2 min and 95uC for 10 min followed by 40 cycles of 95uC for 15 sec and 60uC for 60 sec), using SYBR green as reporter dye (Applied Biosystems Inc.). Serial 2-fold dilutions from a pool of ConA stimulated DA lymphocyte cDNA was used as standard for assessment of relative efficiency of target amplification. We used the DDCT method for relative quantification as the PCR efficiencies between the target(s) and reference gene(s) were relatively equivalent. Relative quantity (RQ) of expression was calculated as 2 2DDCt (DDCt = (target gene -reference gene)average target gene). Primers sequences for the quantitative real time PCR are found in Table S3.

In vitro experiment
Stimulation. For stimulation experiments 1610 6 cells/well were plated in 48 well flat bottom plates (Nunc, Roskilde, Denmark). Cells were stimulated with either: complete media, 0.5 mg/ml Concanavalin A (Sigma-Aldrich) or 25 mg/ml rMOG for 24 hrs at 37uC and 5% CO 2 . Supernatants were collected for IL-2 ELISA. Cells were spun down and used for flow cytometry.

IL-2 Antibody
Enzyme Linked ImmunoSorbent Assays. IL-2 ELISAs were performed according to manufacturers' recommendations (Biosource International, California, USA) using 96-well flat-bottom ELISA plates (Nunc). Supernatants from stimulation assays were collected at 24 hrs. Flow Cytometry. Lymph node cells were washed with cold PBS and resuspended in a further 100 ml of PBS. Cells were first stained for extracellular CD4-PE-Cy5 and CD25-PE (20 min at 4uC; BD Biosciences, San Jose, USA) and thereafter fixed and stained for intracellular Foxp3-APC according to manufacturers' recommendations (eBioscience, San Diego, USA). Staining was visualized on a FACSort (BD, Franklin Lakes, USA) with Cell Quest (version 3.2.1f1, BD) and analyzed using FloJo (version 8.8).  Figure S2 Eae23 regulates EAE in AIL rats although males have low power. Log-likelihood plots of Eae23 in 794 (DAxPV-G.1AV1)G10 AIL rats. Linkage analysis performed in groups stratified according to sex identified two separate QTLs strongly linked to EAE phenotypes in females while males showed one centered QTL below the threshold for significance. To investigate the possibility of sex-specific QTLs or sex interacting with genotype, we included sex as a covariate in the model. Sex was not significantly contributing to the linkage of Eae23 to clinical phenotypes. Males weighed more than females at the age of immunization and the dose used was possibly suboptimal to induce EAE in the male population, which reduced the power to separate the QTLs in the joint analysis. Table S1 Primer sequences were retrieved from Ensembl Genome Database (http://www.ensembl.org v.55). a Positions are given in megabasepair. The skewing of genotype distributions reflect regions of genome that are becoming fixed in the population due to the genetic drift within the breeding couples used (50 per generation). The region between markers D17Got49 and D17Mgh5 has an overrepresentation of DA/DA genotype (,29% in the 7th AIL generation and ,35% in the 10th AIL generation) and marker D17Rat95 has drifted toward PVG/PVG genotype. The minimum number of individuals with a particular genotype was 68 rats, which should be sufficient for linkage. Additionally, genotyping of selected markers within these regions were repeated to ensure that genotypes were correctly scored.   Author Contributions