The Journal of Immunology

High Resolution Mapping of Cia3: A Common Arthritis Quantitative Trait Loci in Different Species1

Xinhua Yu, Haidong Teng, Andreia Marques, Farahnaz Ashgari, and Saleh M. Ibrahim2

Murine collagen induced arthritis (CIA) is a widely used model of rheumatoid arthritis (RA). Identification of CIA susceptibility will aid in the understanding of RA pathogenesis and development of therapeutic targets. This study aims to identify and refine ؋ quantitative trait loci (QTL) controlling CIA. Major CIA clinical traits were evaluated in both (DBA/1 FVB/N) F2 and advanced intercross line (AIL) mice; QTLs were confirmed and refined in AIL. To search for candidate genes, we applied multiple approaches, including expression profiling, identification of nonsynonymous polymorphism, and comparative genomic mapping. We identified six suggestive QTLs controlling CIA clinical traits in the F2 progeny; one of these was confirmed and refined in AIL. This QTL is located on 6 and overlaps with Cia3, which was identified previously. We refined the 2-log support interval of Cia3 into a 5.6 Mb genomic region; 15 of 77 genes are differentially expressed or carry nonsynonymous polymorphisms between two parental strains. The counterpart genomic region of Cia3 on the rat and human genomes are linked to RA. Twenty-nine of 77 genes are located in the arthritis-linked genomic regions of all three species. Five of those 29 genes are differentially expressed or carry nonsynonymous poly- morphisms between parental strains: Timp4, Tmem40, Mbd4, Cacna1c, and Lrtm2. Taken together, we refined Cia3 into a 5.6 Mb genomic region on mouse chromosome 6 and identified candidate genes. This will aid in the search for susceptibility gene(s) controlling arthritis development within Cia3 and its counterpart regions in rat and human genomes. The Journal of Immunology, 2009, 182: 3016–3023.

heumatoid arthritis (RA)3 is a chronic inflammatory auto- defined as a QTL. Therefore, the number of susceptibility genes could immune joint disease influenced by genetic and environ- be much larger than the number of the identified QTLs. On the other R mental factors (1). The genetic contribution to RA suscep- hand, identification of susceptibility genes within the QTLs is still a tibility is estimated to be as much as 60%, of which the HLA DRB1 challenging task, with the exceptions of few genes with very strong is thought to account for 30–50% (2, 3). However, identification effect on the disease, e.g., NCF1 (7). In most cases, a single quanti- of non-MHC RA susceptibility genes has been challenging due to tative trait gene contributes only mildly or moderately to the outcome genetic heterogeneity and incomplete penetrance, as well as the effect of the complex traits. To accelerate progress in the identification of of environmental factors on RA development. Presently, only a few susceptibility genes, several complementary approaches have been genes have been convincingly showed to be associated with RA, in- suggested, such as identification of polymorphisms in coding or reg- cluding PADI4, PTPN22, and CTLA4 (4, 5, 6). Genetic analysis of ulatory region, in vitro functional studies, transgenesis, knock-in mod- well-defined experimental models of autoimmune arthritis provides els, deficiency-complementation testing, mutational analysis, and ho- an alternative strategy to study the genetic basis of RA. A good ex- mology searches (10). In addition, advanced progress in mouse ample is identification of NCF1 as a novel susceptibility gene in au- genetics also accelerates progress in the identification of the quanti- toimmune arthritis (7). Animal models of RA have been used to iden- tative trait genes. For example, genomes of the 16 commonly used tify susceptibility genes, and multiple quantitative trait loci (QTLs) mouse inbred strains have been recently sequenced, and 8.27 million have been identified (see Ref. 8, 9). SNPs have been identified (11). Therefore, polymorphisms within a Despite the many advantages of animal models, identification of QTL identified in those 16 strains or their substrains can be obtained, susceptibility genes in animal models is limited by two factors. On which will considerably help in the identification of candidate genes. one hand, linkage analysis is not as powerful as case-control associ- Previously we performed a genome-wide linkage analysis in a F2 ation studies. As a consequence, a genomic region containing a sus- progeny to identify QTLs controlling collagen-induced arthritis (CIA) ceptibility gene that has very small effect on disease might not reach and generated an advanced intercross line (AIL) to refine those QTLs the significant threshold of the linkage analysis, and thus will not be (12, 13). In the F2 progeny, we identified one QTL with a strong effect on CIA clinical traits, Cia2, that contributes to only 16% of the phe- Section of Immunogenetics, University of Rostock, Germany notype variant (12). This suggests that there could be additional sus- ceptibility gene(s) whose effect is masked by Cia2, and thus failed to Received for publication September 10, 2008. Accepted for publication December 24, 2008. reach a significant threshold. In this study, we investigated six

The costs of publication of this article were defrayed in part by the payment of page genomic regions that showed suggestive level of linkage to CIA in F2 charges. This article must therefore be hereby marked advertisement in accordance mice. We used the AIL to confirm those suggestive QTLs and refine with 18 U.S.C. Section 1734 solely to indicate this fact. the positive(s) into a small genomic region. Also, we investigated the 1 This study was supported by the EU FP6 contract MRTN-CT-2004-005693 candidate genes for one confirmed and refined QTL by defining dif- (EURO-RA). ferentially expressed genes, identifying the nonsynonymous polymor- 2 Address correspondence and reprint requests to Dr. Saleh M. Ibrahim at the current address: Genetics Group, Department of Dermatology, University of Lu¨beck, Ratzeburger phisms, and performing comparative genomic mapping. Allee 160, 23538 Lu¨beck, Germany. E-mail address: [email protected] 3 Abbreviations used in this paper: RA, rheumatoid arthritis; QTL, quantitative Materials and Methods trait loci; CIA, collagen-induced arthritis; AIL, advanced intercross line; chr., Phenotypic traits of CIA chromosome; SNP, single nucleotide polymorphism. ϫ Two mouse populations used in this study, 290 (DBA/1 FVB/N)F2 ϫ Copyright © 2009 by The American Association of Immunologists, Inc. 0022-1767/09/$2.00 mice and 308 (DBA/1 FVB/N)F11/12 AIL mice, were generated

www.jimmunol.org/cgi/doi/10.4049/jimmunol.0803005 The Journal of Immunology 3017

Table I. List of QTLs controlling clinical traits of CIA

Chr. Marker LOD Score Traits Susceptible Allele Overlap with CIA QTLs severity, onset, susceptibility DBA/1 Cia2, Cia4 ءءD2Mit81 10.4 2 severity FVB/N Cia3 ءD6Mit328 2.37 6 severity FVB/N ءD7Mit248 2.12 7 severity DBA/1 Cia8 ءD10Mit261 2.1 10 severity FVB/N Cia40 ءD11Mit126 2.08 11 susceptibility FVB/N ءD18Mit222 2.43 18 onset FVB/N ءD19Mit90 2.66 19

.suggestive ,ء highly significant and ,ءء previously in our laboratory. Detailed information of mice characteris- reevaluated using the scoring system that was applied in AIL mice (13). tics and induction of CIA were described previously (12, 13). Three The onset trait of the F2 and AIL mice were calculated previously (12, clinical traits of CIA were used for linkage analysis: severity, onset, and 13). Susceptibility is a qualitative trait, with a score of 0 and 1 for the susceptibility. The CIA severity (maximal score) in the F2 progeny was healthy and diseased mice, respectively.

Table II. ANOVA analysis of with evidence of linkage in F2 and AIL

F2 mice AIL mice

Chr. Markers Position (Mb) F p value Markersa Position (Mb) F p value

6 D6Mit67 97.7 2.92 0.053000 D6Mit67 97.7 0.63 0.531 D6Mit328 112.7 5.52 0.004400 D6Mit328 112.7 3.56 0.029 D6MIt10 113.2 3.95 0.02 D6Mit329 114.1 3.96 0.02 D6Mit366 115.2 5.96 0.00278 D6Mit115 116.7 8.7 0.00021 rs51294806 117.6 6.12 0.00264 rs6295683 118.5 5.77 0.00345 rs30265977 119.7 2.51 0.082 rs50344715 123.5 1.12 0.327 D6Mit335 127.5 2.78 0.0638 D6Mit14 145.6 1.35 0.261000 D6Mit14 145.6 0.47 0.624 7 D7Mit228 47.3 2.69 0.070900 D7Mit228 47.3 0.94 0.392 D7Mit229 52.9 2.13 0.12 D7Mit193 57 1.66 0.191 D7Mit83 59.1 1.39 0.249 D7Mit295 63.6 1.52 0.219 D7Mit88 67.3 1.49 0.226 D7Mit248 73 5.17 0.006200 D7Mit248 73 1.94 0.144 D7Mit122 82.4 0.76 0.467 D7Mit350 90.7 2.59 0.076000 D7Mit350 90.7 0.96 0.382 D7Mit183 101.6 0.58 0.561 D7Mit323 108 1.26 0.285 10 D10Mit20 66.5 2.94 0.054000 D10Mit20 66.5 1.53 0.217 D10Mit32 69.1 1.91 0.148 D10Mit186 75.3 1.08 0.341 D10Mit174 75.7 0.54 0.582 D10Mit132 83.6 0.56 0.569 D10Mit261 85.1 4.88 0.008100 D10Mit261 85.1 1.47 0.231 D10Mit94 87.7 0.32 0.722 D10Mit161 90.2 0.83 0.436 D10Mit41 93.7 0.12 0.887 D10Mit96 99.1 3.24 0.041000 D10Mit96 99.1 0.97 0.378 D10Mit70 103.5 0.348 0.706 11 D11Mit285 89.7 3.21 0.041700 D11Mit285 89.7 0.47 0.493 D11Mit70 93.9 1.75 0.175 D11Mit289 94.7 0.84 0.432 D11Mit145 97.5 3.2 0.041 D11Mit126 103.7 5 0.007300 D11Mit126 103.7 0.59 0.551 D11Mit58 104.4 0.81 0.447 rs27054829 105.5 0.86 0.425 rs27004424 106.4 0.14 0.873 rs27010185 107.4 2.54 0.08 D11Mit100 110 3.41 0.0344 D11Mit214 114.9 4.68 0.009900 D11Mit214 114.9 1.7 0.183 18 D18Mit67 12.1 0.93 0.396 D18Mit222 14.7 5.63 0.003900 D18Mit222 14.7 0.91 0.404 D18Mit230 17.8 1.65 0.193 D18Mit22 25.1 1.2 0.302 D18Mit12 36 2.75 0.065000 D18Mit12 36 0.6 0.548 19 D19Mit40 25.4 1.74 0.176 D19Mit46 32.7 1.21 0.297 D19Mit88 37.3 5.32 0.005300 D19Mit88 37.3 1.62 0.197 D19Mit90 42.2 6.58 0.001600 D19MIt11 42 1.18 0.308 D19Mit38 47 0.09 0.91 rs31054271 53.1 0.94 0.389 D19Mit84 56 1.27 0.28 rs37383437 57.3 1.83 0.161 D19Mit71 59.6 6.35 0.001900 D19Mit71 59.6 4.57 0.011 rs46580758 60.9 2.83 0.06

a Markers showing significant linkage are marked in bold. 3018 FINE-MAPPING ARTHRITIS QTL

Markers and genotyping

To confirm and refine the suggestive QTLs identified in F2 progeny, we genotyped the genomic regions containing those QTLs in AIL mice. Be- cause the confidence intervals of QTLs in this AIL range from 4 to 12 Mb (13, 14), we first selected genetic markers covering those genomic regions with intermarker distance of ϳ5 Mb. Then, we increased the marker den- sity for the genomic regions showing evidence of linkage ( p Ͻ 0.05). Genotyping of AIL mice was performed on DNA extracted from tail tips using PCR amplification for microsatellite markers as described previously (12) or by a PCR-RFLP method for SNP markers. We genotyped 308 AIL mice with 38 markers and we included 22 markers genotyped previously (14). In total, 60 genetic markers (50 microsatellite markers and 10 SNP markers) in six genomic regions were used for analysis. Genotypes of all nonsynonymous SNPs in FVB/NJ and DBA/2J strain and genotypes of part of the nonsynonymous SNPs in DBA/ 1J strain were retrieved from the Mouse Phenome Database (http:// phenome.jax.org/pub-cgi/phenome/mpdcgi?rtn ϭ snps/door). We geno- typed the DBA/1J strains with the nonsynonymous SNPs located in the QTL confidence interval by sequencing. We designed primers to amplify the genomic fragment comprising a SNP and sequenced the PCR product directly. Gene expression profiling Previously, we detected gene expression profiling in lymph nodes of DBA/1 and FVB/N mice on day 0 (before immunization), day 10, day 35 (onset phase), as well as day 95 (chronic phase) after CIA induction (15). In this study, we analyzed the gene expression profiling in joint of DBA/1 and FVB/N mice on days 0 and 35 as well as in the thymus of the two strains on day 0. Each group contained three mice. Analysis of gene ex- pression was conducted using MOE 430A array (Affymetrix), interrogating more than 22,000 genes according to procedures described previously (15, 16). The normalized expression values were imported to and analyzed by Affymetrix dCHIP software (17). Differentially expressed genes were iden- tified by defining the following filtering criteria in the dCHIP software: 1) The fold change between the group means exceeded 2-fold; 2) The abso- lute difference between the two groups exceeded 100; and 3) The p-value threshold of the unpaired t test was 0.05. The false discovery rate was established with permutation test for each pairwise comparison to estimate the proportion of false-positive genes. Linkage analysis All linkage analyses were performed using QTX Map manager software (18). The physical positions of the loci were obtained from Ensembl (http:// www.ensembl.org). The suggestive and significance linkage threshold val- ues were determined by permutation tests (n ϭ 500). In accordance with the suggestion of Lander and Botstein (19), the confidence interval was defined as the distance between points on each side of the peak of each FIGURE 1. Cia3 in F2 and AIL mice. Development of CIA in F2 (A) QTL at which the LOD score falls by 2. and in AIL (B). Mice were divided into three groups according to the genotype of peak markers (D6Mit328 in F and D6Mit115 in AIL) of Comparative genomic mapping 2 Cia3, where aa, ab, and bb stand for FVB/N homozygous, heterozygous, The comparative mapping was performed using HomoloGene orthology and DBA/1 homozygous alleles. C, Log-likelihood plot showing the predictions (http://www.ncbi.nlm.nih.gov/projects/homology/maps/) among relationship between Cia3 and arthritis clinical traits. The horizontal mice, rats, and humans. The genomic region of the confidence interval of the line indicates the significant threshold defined by the permutation test. mouse CIA QTL and its counterparts on rat and human genomes were The filled bar indicates the 2-log confidence interval. The genetic mark- analyzed. The confidence intervals of rat arthritis QTLs are retrieved from ers used in study are indicated according to their physical positions. the original reports, and the confidence intervals of human genomic regions were artificially defined as 10 Mb at both sides of the microsatellite mark- ers that are linked to RA. tive scoring system. In addition to Cia2, six suggestive QTLs were identified on chromosome 6, 7, 10, 11, 18, and 19, with LOD Results scores ranging from 2.08 to 2.66. Four of these suggestive QTLs Suggestive QTLs controlling CIA clinical traits in F2 mice control CIA severity, and two control CIA onset and susceptibility, Previously, we performed a genome-wide linkage analysis using respectively (Table I). Interestingly, three of these QTLs overlap ϫ (DBA/1 FVB/N) F2 progeny to identify QTLs controlling CIA with previously identified mouse CIA QTLs: they are Cia3 on chr. (12). A genomic region on chromosome (chr.) 2 was significantly 6, Cia8 on chr. 10, and Cia40 on chr. 11 (12, 20–23). These over- linked to CIA clinical traits, e.g., severity, onset, and susceptibility. laps support the idea that susceptibility genes could be located in This genomic region was named Cia2 (12, 20, 21). No other sig- those regions. nificant QTLs controlling CIA clinical traits were identified in the ϳ Confirming suggestive QTLs in AIL F2 progeny; however, the Cia2 contributes only 16% of the clin- ical variation. This suggests that some susceptibility genes with To confirm the suggestive QTLs, we used an AIL that we gener- relatively small effect on CIA failed to reach the significance link- ated previously (13). We genotyped 308 AIL mice with 60 genetic age threshold due to the masking effect of Cia2. To test this hy- markers covering the six QTLs. We then performed ANOVA anal- pothesis, we reanalyzed the F2 mice with suggestive significance ysis and QTL Linkage analysis for each marker. Table II summa- thresholds. Also, we revaluated CIA severity using a more sensi- rizes the results of the ANOVA analysis. Among the six suggestive The Journal of Immunology 3019

Table III. List of nonsynonymous gene polymorphism with Cia3a

SNP Position (Mb) Allele Gene SAP DBA/2 FVB/N DBA/1

rs13478988 115.5 C/G 2510049J12Rik R101G C G C rs31495179 115.7 C/T Tmem40 T118A T C T rs31498257 115.7 A/T Tmem40 Q31H A T A rs31498687 115.7 A/G Tmem40 S3L G A G rs30840549 115.8 C/T Mbd4 N128D T C T rs30121304 116.4 C/T Alox5 I645V T C T rs31551252 116.5 A/G Olfr212 Q15R A G A rs31551918 116.5 G/T Olfr212 V56L T G T rs31553514 116.5 A/G Olfr212 Q239R A G A rs31549850 116.5 G/T Olfr213 V56L T G T rs37185903 116.5 C/T Olfr215 I213V T C T rs36953372 117.4 C/G LOC100043777 F2L C G C rs37558791 117.8 C/G Zfp239 T56S C G C rs37667924 117.8 A/C Zfp239 R127S C A C rs31557206 118.4 A/C Bms1 G637V A CC rs31557210 118.4 C/T Bms1 E550G T C C rs31557212 118.4 C/G Bms1 A520G C G C rs31563226 118.4 A/G Bms1 A364V G A G rs31562616 118.4 A/T Zfp248 T306S T T A rs31574016 118.5 G/T Ankrd26 E1664D G T G rs31574018 118.5 C/G Ankrd26 P1657A C G C rs31570654 118.5 C/T Ankrd26 Y1509C T C T rs31571560 118.5 A/G Ankrd26 A1450V G A G rs31573792 118.5 G/T Ankrd26 Q1363H G T G rs31572531 118.5 C/T Ankrd26 R848G C T C rs30363420 118.5 C/T Ankrd26 T784A T C T rs31576356 118.5 C/T Ankrd26 N449S C T C rs29873127 118.5 A/T Ankrd26 T425S A T A rs31578998 118.5 C/G Ankrd26 H398D G C G rs31579689 118.5 C/T Ankrd26 I217M C T C rs31580309 118.5 C/T Ankrd26 N113S C T C rs31576214 118.5 C/G Ankrd26 H36Q G C G rs31579954 118.6 G/T Cacna1c N1769T G G T rs31661638 119.3 G/T Lrtm2 I35L T G T

a Data obtained from the Mouse Phenome Database.

QTLs, only the QTL on chr. 6 (Cia3) showed significant linkage to refined Cia3 into a 5.6 Mb genomic region with flanking markers CIA. The peak marker of Cia3 is D6Mit115 (F ϭ 8.7, p ϭ of D6Mit329 and rs30265977 (Fig. 1C).

0.00021), which is 3.9 Mb away from the peak marker in F2 mice. The other five genomic regions failed to reach a significant level of Nonsynonymous polymorphism within Cia3 linkage, although the peak markers on chr. 11 and chr. 19 showed A quantitative trait gene polymorphism is either a nonsynony- slight linkage ( p ϭ 0.034 and p ϭ 0.011, respectively). Therefore, mous polymorphism changing protein structure or a regulatory Cia3 has been confirmed as a CIA QTL in the DBA/1 ϫ FVB/N sequence variation affecting gene expression. Therefore, iden- cross. tification of the nonsynonymous polymorphisms between pa- rental strains could aid in the search for candidate genes. How- Fine mapping of Cia3 ever, Cia3 is located in a high gene-density genomic region, ϫ Cia3 was originally identified in (DBA/1 SWR) F2 mice, with containing 77 genes. Sequencing all the coding region of those the peak marker of D6Mit10 located at 113.2 Mb on chr. 6 (21). genes is time and resource consuming. Fortunately, 16 common Our study confirmed this QTL in an independent cross. The peak mouse inbred strains have been recently sequenced and 8.4 mil- markers of Cia3 in our F2 and AIL are D6Mit328 (112.7 Mb) and lion SNPs have been identified among them (11). FVB/NJ and

D6Mit115 (116.6 Mb), respectively. In both F2 and AIL, the DBA/2J strains are among those 16 strains. DBA/1J and FVB/N allele enhance the disease in an additive manner (Fig. 1, A DBA/2J are substrains, and the confidence interval of Cia3 is an and B). The confidence interval of Cia3 in both (DBA/1 ϫ SWR) identical by descent chromosomal region between the two sub- ϫ Ͼ F2 and (DBA/1 FVB/N) F2 mice are 40 Mb. Using AIL, we strains. Therefore, searching for nonsynonymous SNPs between

Table IV. List of differentially expressed genes within Cia3

Thymus LN Joint

Gene Position (Mb) Probe ID Day 0 Day 0 Day 10 Day 35 Day 95 Day 0 Day35

Timp4 115,2 1423405_at ϪϪϪϪUp (2.7) Up (2.1) Ϫ Cacna1c 118,5 1421297_a_at Down (Ϫ4.0) ϪϪ Ϫ Ϫ ϪϪ

Ϫ, not differentially expressed; Up, with higher expression in DBA/1 strain; Down, with lower expression in DBA/1 strain; ( ), fold change. 3020 FINE-MAPPING ARTHRITIS QTL

FIGURE 2. Comparative mapping of genomic regions containing Cia3 among mouse, rat, and human. The comparative maps are calculated using HomoloGene orthology predictions (http://www.ncbi.nlm.nih.gov/projects/homology/maps/) for the mouse Cia3. The physical positions of the chromosome are presented in Mb. The confidence intervals of arthritis QTLs and human genomic regions linked to the RA are present as black bars. The confidence intervals of rat Pia7 and Cia13 are from the original reports, and the confidence intervals of the are artificially defined as 10 Mb at both side of the microsatellite markers that are linked to RA. Fifty-seven known genes within Cia3 are depicted in the figure according to their physical positions. The gray areas indicate the homologous region between mouse and rat as well as between mouse and human.

DBA/2J and FVB/N in this region could help to identify the gene expression in the thymus and joints of both strains. In total, polymorphisms between DBA/1J and FVB/N strains. In total, 1312 genes were differentially expressed between DBA/1 and 32 nonsynonymous SNPs were found between DBA/2J and FVB/N strains in one or more tissues (Supplementary Table I).4 FVB/NJ strains. Genotyping DBA/1J strain with those SNPs Two differentially expressed genes are located within the confi- showed that DBA/1J shared the same alleles with DBA/2J strain dence interval of Cia3. One is tissue inhibitor of metalloprotein- in 30 of 32 SNPs. An additional two nonsynonymous SNPs that ases 4 (Timp4), with higher expression in DBA/1 strain than in are not polymorphic between FVB/N and DBA/2J strains were FVB/N strain in lymph nodes during chronic phase and in the joint found to be polymorphic between DBA/1 and FVB/N strains. before immunization. The other gene is voltage-dependent L-type Taken together, 32 nonsynonymous SNPs in 14 genes were calcium channel subunit ␣-1C (Cacna1c), with lower expression in identified between DBA/1 and FVB/N strains (Table III). DBA/1J than FVB/NJ in the thymus (Table IV).

Gene expression profiling Comparative genomic mapping of Cia3 When a quantitative trait gene polymorphism is a sequence vari- When a genomic region and its counterparts in multiple species are ation regulating the expression of a gene, the gene should be dif- identified as QTLs controlling diseases, it indicates that a common ferentially expressed between parental strains in disease-related susceptibility gene might exist in multiple species. Based on this tissue in a certain phase of the disease. Therefore, defining the hypothesis, comparative mapping among multiple species could be gene expression profile in disease related tissues could aid in the performed to refine a QTL and to select candidate genes. For the search for candidate genes. Previously, we performed gene expres- refined 5.6 Mb confidence interval of mouse Cia3, the counterpart sion profiling on lymph nodes of DBA/1 and FVB/N strains at four disease phases during the development of CIA (15). To complete the gene expression profiling in CIA related tissues, we detected 4 The online version of this article contains supplementary material. The Journal of Immunology 3021

Table V. List of genes located in arthritis QTLs in mouse, rat, and human

Mouse Position (Mb) Gene Description Rat Position (Mb) Human Position (Mb)

Chr. 6 114.1 Slc6a11 solute carrier family 6, member 11 Chr. 4 150.1 Chr. 3 10.8 114.2 Slc6a1 solute carrier family 6, member 1 150.2 11.0 114.3 Hrh1 histamine receptor H 1 150.4 11.2 114.6 Atg7 autophagy-related 7 (yeast) 150.8 11.3 114.8 Vgll4 vestigial like 4 (Drosophila) 151.0 11.6 115.0 1500001M20Rik RIKEN cDNA 1500001M20 gene 151.1 11.8 115.1 Syn2 synapsin II 151.3 12.0 115.2a Timp4 tissue inhibitor of metalloproteinase 4 151.4 12.2 115.4 Pparg peroxisome proliferator activated receptor gamma 151.6 12.3 115.5 Tsen2 tRNA splicing endonuclease 2 homolog 151.7 12.5 115.6 Mkrn2 makorin, ring finger protein, 2 151.7 12.6 115.6 Raf1 v-raf-leukemia viral oncogene 1 151.8 12.6 115.7 Tmem40 transmembrane protein 40 151.9 12.8 115.7 Cand2 cullin-associated and neddylation-dissociated 2 151.9 12.8 115.8 BC060267 cDNA sequence BC060267 151.9 130.6 115.8 Mbd4 methyl-CpG binding domain protein 4 152.0 130.6 115.8 Ift122 intraflagellar transport 122 homolog 152.0 130.6 115.9 Rho rhodopsin 152.1 130.7 115.9 H1foo H1 histone family, member O, oocyte-specific 152.1 130.7 115.9 Plxnd1 plexin D1 152.1 130.8 116.0 Tmcc1 transmembrane and coiled coil domains 1 152.1 130.8 119.1 Dcp1b DCP1 decapping enzyme homolog b 155.5 Chr. 12 1.9 118.5 Cacna1c calcium channel, voltage-dependent, L type, 154.9 2.0 alpha 1C subunit 119.2 Cacna2d4 calcium channel, voltage-dependent, ␣2/␦ subunit 4 155.5 1.8 119.3 Adipor2 adiponectin receptor 2 155.7 1.7 119.4 Wnt5b wingless-related MMTV integration site 5B 155.7 1.6 119.3 Lrtm2 leucine-rich repeats and transmembrane domains 2 155.6 1.8 119.4 Fbxl14 F-box and leucine-rich repeat protein 14 155.8 1.5 119.5 Erc1 ELKS/RAB6-interacting/CAST family member 1 155.9 1.0

a Genes carrying non-synonymous polymorphism(s) or differentially expressed between DBA/1 and FVB/N strains are written in bold letters.

on the rat genome isa6Mb(150–156 Mb) genomic region on cates that the two genomic regions might contain small effect CIA chromosome 4. The peaks of two rat arthritis QTLs, Pia7 and susceptibility genes which remain to be confirmed. Cia13, are located in this region (24, 25) (Fig. 2). The counterpart Cia3 was identified in a previous study with an independent ϫ of Cia3 on the human genome is mapped to three chromosomes: population of (DBA/1 SWR) F2 progeny (21). In our study, the , 10, and 12. The counterpart genomic regions on allele from the resistant strain, FVB/N, enhanced susceptibility to ϫ chromosome 3 and 12 have been showed to be linked to RA (26– CIA. The susceptible allele of Cia3 in (DBA/1 SWR) F2 prog- 28) (Fig. 2). When comparative mapping was performed, 29 of the eny was not shown. However, SWR mice with Cia2 congenic 77 genes within the Cia3 confidence interval were presentein fragment from DBA/1 strain showed higher susceptibility to CIA genomic regions linked to arthritis in all three species (Table V). than DBA/1 mice, indicating that the SWR strain carries CIA sus-

Among the 29 genes, five genes carry nonsynonymous polymor- ceptibility allele(s) (21). In both F2 and AIL mice, Cia3 affects phism or are differentially expressed between DBA/1 and FVB/N CIA severity, onset, and susceptibility. However, the magnitudes strains. These are Timp4, Tmem40, Mbd4, Cacna1c, and Lrtm2. of the effects on individual traits vary between F2 and AIL, with

Recently, two genome-wide association studies have been reported highest effect on severity in F2 and highest effect on susceptibility (29, 30), providing an additional novel resource for comparative in AIL. genome mapping. In these two published reports, we looked into We refined Cia3 into a 5.6 Mb genomic region which con- the association of SNPs located within counterpart genomic re- tains 77 genes, including 57 known genes and 20 predicted or gions of Cia3 on human genome with RA. However, no significant hypothetical genes. This relatively small genomic region al- association was observed. lowed us to realistically search for the candidate genes. To se- lect the candidate genes, we performed gene expression profil- Discussion ing in disease-related tissues at different phases of the disease. The aim of a significance threshold applied to QTLs is to decrease We also searched for the nonsynonymous polymorphisms in the false-positives to a reasonable level, with the drawback of in- coding sequence. In total, we identified 15 genes that are either creasing the false negatives (19). Therefore, QTLs identified in a differentially expressed or carry nonsynonymous polymor-

F2 progeny need to be confirmed in other populations, e.g., con- phism(s) between the parental strains. These genes could be genic strains and AIL. In this study, we used an AIL progeny to considered as candidate genes for further studies. However, this confirm and refine suggestive QTLs identified in F2 progeny. One gene list is not conclusive for two reasons. First, we identified of six suggestive QTLs, Cia3, was confirmed in the AIL and re- nonsynonymous polymorphisms between FVB/N and DBA/1 fined into a 5.6 Mb genomic region. This is one practical example strains by initially comparing FVB/N and DBA/2 strains. of a false negative QTL in F2 mice. Besides Cia3, two other sug- DBA/1 and DBA2 are substrains and the confidence interval of gestive QTLs, on chr. 11 (Cia40) and on chr. 19, warrant discus- Cia3 is an identical by descent region. Sequences in the sion. Although not significant, there is slight linkage with the same genomic region of the two substrains should be identical, with susceptibility alleles in AIL as in F2 (data not shown). This indi- exception of polymorphisms occurring after the separation of 3022 FINE-MAPPING ARTHRITIS QTL the substrains ϳ80 years ago (31). Secondly, gene expression 3. Wordsworth, B. P., J. S. Lanchbury, L. I. Sakkas, K. I. Welsh, G. S. Panayi, and profiling was performed with the MOE 430A array, which has J. I. Bell. 1989. HLA-DR4 subtype frequencies in rheumatoid arthritis indicate that DRB1 is the major susceptibility locus within the HLA class II region. Proc. ϳ22,000 genes. Regarding the 77 genes within Cia3, 28 were Natl. Acad. Sci. USA 86: 10049–10053. not included in MOE 430A array, and thus their expression 4. Suzuki, A., R. Yamada, X. Chang, S. Tokuhiro, T. Sawada, M. Suzuki, M. Nagasaki, M. Nakayama-Hamada, R. Kawaida, M. Ono, et al. 2003. Func- could not be detected. The 28 genes contain 20 predicted or tional haplotypes of PADI4, encoding citrullinating enzyme peptidylarginine de- hypothetical genes and eight known genes. The eight known iminase 4, are associated with rheumatoid arthritis. Nat. Genet. 34: 395–402. genes are D830050J10Rik, olfactory receptor 211, olfactory re- 5. Begovich, A. B., V. E. Carlton, L. A. Honigberg, S. J. Schrodi, A. P. Chokkalingam, H. C. Alexander, K. G. Ardlie, Q. Huang, A. M. Smith, ceptor 212, olfactory receptor 2113, olfactory receptor 214, ol- J. M. Spoerke, et al. 2004. A missense single-nucleotide polymorphism in a gene factory receptor 215, 4933440N22Rik, and 1700069P05Rik encoding a protein tyrosine phosphatase (PTPN22) is associated with rheumatoid (supplementary Table II). Taken together, the 15 genes on the arthritis. Am. J. Hum. Genet. 75: 330–337. 6. Plenge, R. M., L. Padyukov, E. F. Remmers, S. Purcell, A. T. Lee, E. W. Karlson, list are promising candidate genes for Cia3, but there is still a F. Wolfe, D. L. Kastner, L. Alfredsson, D. Altshuler, et al. 2005. Replication of possibility, although very small, that the susceptibility gene is putative candidate-gene associations with rheumatoid arthritis in Ͼ4,000 samples from North America and Sweden: association of susceptibility with PTPN22, outside of this list. CTLA4, and PADI4. Am. J. Hum. Genet. 77: 1044–1060. Making use of the abundant resource of QTLs in RA and its 7. Olofsson, P., J. Holmberg, J. Tordsson, S. Lu, B. Akerstrom, and R. Holmdahl. animal models, we performed comparative genomic mapping 2003. Positional identification of Ncf1 as a gene that regulates arthritis severity in rats. Nat. Genet. 33: 25–32. for Cia3. Based on the hypothesis that the same genomic region 8. Holmdahl, R. 2006. Dissection of the genetic complexity of arthritis using animal in multiple species linked to the disease could contain a com- models. Immunol. Lett. 103: 86–91. mon susceptibility gene, we shortened the Cia3 gene list from 9. Ibrahim, S. M., and X. Yu. 2006. Dissecting the genetic basis of rheumatoid arthritis in mouse models. Curr. Pharm. Des. 12: 3753–3759. 77 to 29 genes. Moreover, when we merged this 29-genes list 10. Abiola, O., J. M. Angel, P. Avner, A. A. Bachmanov, J. K. Belknap, B. Bennett, with the list of 15 genes that carry nonsynonymous polymor- E. P. Blankenhorn, D. A. Blizard, V. Bolivar, G. A. Brockmann, et al. 2003. The phism or are differentially expressed between DBA/1 and nature and identification of quantitative trait loci: a community’s view. Nat. Rev. Genet. 4: 911–916. FVB/N strains, five genes appear to be of particular interest. 11. Frazer, K. A., E. Eskin, H. M. Kang, M. A. Bogue, D. A. Hinds, E. J. Beilharz, These are Timp4, Tmem40, Mbd4, Cacna1c, and Lrtm2. The R. V. Gupta, J. Montgomery, M. M. Morenzoni, G. B. Nilsen, et al. 2007. A Timp4 gene is a member of gene family of issue inhibitors of sequence-based variation map of 8.27 million SNPs in inbred mouse strains. Nature 448: 1050–1053. matrix metalloproteinases (32). The Mbd4 gene encode a pro- 12. Bauer, K., X. Yu, P. Wernhoff, D. Koczan, H. J. Thiesen, and S. M. Ibrahim. tein containing a methyl-CpG binding domain and can enzy- 2004. Identification of new quantitative trait loci in mice with collagen-induced arthritis. Arthritis Rheum. 50: 3721–3728. matically remove thymine (T) or uracil (U) from a mismatched 13. Yu, X., K. Bauer, P. Wernhoff, D. Koczan, S. Moller, H. J. Thiesen, and CpG site in vitro (33). The Cacna1c gene encode sa protein S. M. Ibrahim. 2006. Fine mapping of collagen-induced arthritis quantitative trait comprising approximately one third of the primary structure of loci in an advanced intercross line. J. Immunol. 177: 7042–7049. ␣ 14. Yu, X., K. Bauer, P. Wernhoff, and S. M. Ibrahim. 2007. Using an advanced carboxyl-terminal of the L-type calcium channel which is sen- intercross line to identify quantitative trait loci controlling immune response dur- sitive to limited posttranslational processing (34). The functions ing collagen-induced arthritis. Genes Immun. 8: 296–301. of Tmem40 and Lrtm2 however, are not clear. Among the five 15. Yu, X., K. Bauer, D. Koczan, H. J. Thiesen, and S. M. Ibrahim. 2007. Combining global genome and transcriptome approaches to identify the candidate genes of genes, Timp4 has been showed to be involved in arthritis de- small-effect quantitative trait loci in collagen-induced arthritis. Arthritis Res. velopment. Previous reports showed that expression of Timp4 in Ther. 9: R3. cartilage is decreased in osteoarthritic patients as compared 16. Ibrahim, S. M., D. Koczan, and H. J. Thiesen. 2002. Gene-expression profile of collagen-induced arthritis. J. Autoimmun. 18: 159–167. with controls (35, 36), indicating that expression level of Timp4 17. Li, C., and W. H. Wong. 2001. Model-based analysis of oligonucleotide arrays: might play a role in cartilage destruction. In line with the evi- expression index computation and outlier detection. Proc. Natl. Acad. Sci. USA 98: 31–36. dence from the human studies, FVB/N strain carrying the risk 18. Manly, K. F., R. H. Cudmore, Jr., and J. M. Meer. 2001. Map Manager QTX, allele of Cia3 shows lower expression of the Timp4 gene than cross-platform software for genetic mapping. Mamm. Genome. 12: 930–932. DBA/1 strain. 19. Lander, E. S., and D. Botstein. 1989. Mapping mendelian factors underlying quantitative traits using RFLP linkage maps. Genetics 121: 185–199. In conclusion, we confirmed and refined Cia3 into a 5.6 Mb 20. Johansson, A. C., M. Sundler, P. Kjellen, M. Johannesson, A. Cook, genomic region on mouse chr. 6. With gene expression profil- A. K. Lindqvist, B. Nakken, A. I. Bolstad, R. Jonsson, M. arcon-Riquelme, and ing, sequencing, and comparative mapping, we selected a set of R. Holmdahl. 2001. Genetic control of collagen-induced arthritis in a cross with NOD and C57BL/10 mice is dependent on gene regions encoding complement putative candidate genes. This aids in the search for suscepti- factor 5 and Fc␥RIIb and is not associated with loci controlling diabetes. Eur. bility gene(s) within Cia3 and its counterpart rat QTLs (Pia7 J. Immunol. 31: 1847–1856. and Cia13) as well as susceptibility genes of RA in humans. 21. McIndoe, R. A., B. Bohlman, E. Chi, E. Schuster, M. Lindhardt, and L. Hood. 1999. Localization of non-Mhc collagen-induced arthritis susceptibility loci in Consequently, our future studies will mainly focus on associa- DBA/1j mice. Proc. Natl. Acad. Sci. USA 96: 2210–2214. tion studies in RA case-control cohorts. Functional polymor- 22. Liljander, M., M. A. Sallstrom, S. Andersson, A. Andersson, R. Holmdahl, and R. Mattsson. 2006. Identification of collagen-induced arthritis loci in aged mul- phisms in putative candidate genes will be investigated. In ad- tiparous female mice. Arthritis Res. Ther. 8: R45. dition, Cia3 congenic mice will be generated to functionally 23. Yang, H. T., J. Jirholt, L. Svensson, M. Sundvall, L. Jansson, U. Pettersson, and explore the role of the susceptibility gene in the development of R. Holmdahl. 1999. Identification of genes controlling collagen-induced arthritis in mice: striking homology with susceptibility loci previously identified in the rat. arthritis. J. Immunol. 163: 2916–2921. 24. Olofsson, P., J. Holmberg, U. Pettersson, and R. Holmdahl. 2003. Identification Acknowledgments and isolation of dominant susceptibility loci for pristane-induced arthritis. J. Im- munol. 171: 407–416. We thank Ilona Klamfuss for taking care of the animals. 25. Griffiths, M. M., J. Wang, B. Joe, S. Dracheva, Y. Kawahito, J. S. Shepard, V. R. Reese, S. Call-Vining, A. Hashiramoto, G. W. Cannon, et al. 2000. Iden- Disclosures tification of four new quantitative trait loci regulating arthritis severity and one The authors have no financial conflict of interest. new quantitative trait locus regulating autoantibody production in rats with col- lagen-induced arthritis. Arthritis Rheum. 43: 1278–1289. 26. Jawaheer, D., M. F. Seldin, C. I. Amos, W. V. Chen, R. Shigeta, C. Etzel, References A. Damle, X. Xiao, D. Chen, R. F. Lum, et al. 2003. Screening the genome for 1. Lynn, A. H., C. K. Kwoh, C. M. Venglish, C. E. Aston, and A. Chakravarti. rheumatoid arthritis susceptibility genes: a replication study and combined anal- 1995. Genetic epidemiology of rheumatoid arthritis. Am. J. Hum. Genet. 57: ysis of 512 multicase families. Arthritis Rheum. 48: 906–916. 150–159. 27. Cornelis, F., S. Faure, M. Martinez, J. F. Prud’homme, P. Fritz, C. Dib, H. Alves, 2. MacGregor, A. J., H. Snieder, A. S. Rigby, M. Koskenvuo, J. Kaprio, K. Aho, and P. Barrera, N. de Vries, A. Balsa, et al. 1998. New susceptibility locus for rheu- A. J. Silman. 2000. Characterizing the quantitative genetic contribution to rheu- matoid arthritis suggested by a genome-wide linkage study. Proc. Natl. Acad. Sci. matoid arthritis using data from twins. Arthritis Rheum. 43: 30–37. USA 95: 10746–10750. The Journal of Immunology 3023

28. Jawaheer, D., M. F. Seldin, C. I. Amos, W. V. Chen, R. Shigeta, J. Monteiro, the synapsin 2 gene and tissue distribution of the mRNA. Biochim. Biophys. Acta M. Kern, L. A. Criswell, S. Albani, J. L. Nelson, et al. 2001. A genomewide 1577: 45–52. screen in multiplex rheumatoid arthritis families suggests genetic overlap with 33. Millar, C. B., J. Guy, O. J. Sansom, J. Selfridge, E. MacDougall, B. Hendrich, other autoimmune diseases. Am. J. Hum. Genet. 68: 927–936. P. D. Keightley, S. M. Bishop, A. R. Clarke, and A. Bird. 2002. Enhanced CpG 29. The Wellcome Trust Case Control Consortium. 2007. Genome-wide association mutability and tumorigenesis in MBD4-deficient mice. Science 297: 403–405. study of 14,000 cases of seven common diseases and 3,000 shared controls. 34. Mikala, G., I. Bodi, U. Klockner, M. Varadi, G. Varadi, S. E. Koch, and Nature 447: 661–678. A. Schwartz. 2003. Characterization of auto-regulation of the human cardiac ␣1 30. Plenge, R. M., M. Seielstad, L. Padyukov, A. T. Lee, E. F. Remmers, B. Ding, subunit of the L-type calcium channel: importance of the C-terminus. Mol. Cell A. Liew, H. Khalili, A. Chandrasekaran, L. R. Davies, et al. 2007. TRAF1–C5 as Biochem. 250: 81–89. a risk locus for rheumatoid arthritis: a genomewide study. N. Engl. J. Med. 357: 35. Kevorkian, L., D. A. Young, C. Darrah, S. T. Donell, L. Shepstone, S. Porter, 1199–1209. S. M. Brockbank, D. R. Edwards, A. E. Parker, and I. M. Clark. 2004. Expression 31. Beck, J. A., S. Lloyd, M. Hafezparast, M. Lennon-Pierce, J. T. Eppig, profiling of metalloproteinases and their inhibitors in cartilage. Arthritis Rheum. M. F. Festing, and E. M. Fisher. 2000. Genealogies of mouse inbred strains. Nat. 50: 131–141. Genet. 24: 23–25. 36. Clements, D. N., S. D. Carter, J. F. Innes, W. E. Ollier, and P. J. Day. 2006. 32. Rahkonen, O. P., I. M. Koskivirta, S. M. Oksjoki, E. Jokinen, and E. I. Vuorio. Analysis of normal and osteoarthritic canine cartilage mRNA expression by 2002. Characterization of the murine Timp4 gene, localization within intron 5 of quantitative polymerase chain reaction. Arthritis Res. Ther. 8: R158.