Identification of two novel female-specific non–major histocompatibility complex loci regulating collagen-induced arthritis severity and chronicity, and evidence of epistasis




To identify additional sex-specific and epistatic quantitative trait loci (QTL) regulating collagen-induced arthritis (CIA) severity overall, as well as within different stages during the disease course, in an intercross between major histocompatibility complex–identical inbred rat strains DA/Bkl (susceptible) and ACI/Hsd (resistant).


Arthritic male (DA × ACI)F2 intercross offspring (n = 143) were analyzed separately from the females (n = 184). Phenotypic extremes (maximum arthritis scores [MAS]) were genotyped and used for QTL analysis. All 327 rats were genotyped with the simple sequence-length polymorphism (SSLP) markers closest to the peak of Cia7 and Cia10, the major loci previously identified in this intercross, and with SSLPs covering chromosomes 12 and 18. Phenotypes studied were disease onset, arthritis severity scores on days 14–39, MAS, mean and cumulative arthritis scores, delayed-type hypersensitivity, and antibody responses to rat type II collagen.


A new female-specific arthritis-severity recessive locus was identified on rat chromosome 12 (Cia25), with a maximum effect observed on day 28 (logarithm of odds [LOD] 4.7). The homozygous DA genotype at Cia25 was associated with a 45% higher median arthritis score in females. Sequencing analyses of the Cia25 candidate gene Ncf1 revealed polymorphisms between DA and ACI. The previously identified locus, Cia10, was found to be male-specific. A 2-locus interaction model analysis identified a novel recessive chromosome 18 QTL, Cia26, which was dependent on Cia7, with its maximum effect observed at later stages during the disease course (peak LOD score of 3.6 for arthritis scores on day 39).


This study identified 2 novel female-specific loci, and 1 male-specific locus. Cia25 regulates MAS and disease severity during the mid-to-late stages of the disease course and may be accounted for by Ncf1 polymorphisms. Cia26 is in epistasis with Cia7 and regulates later stages of disease, suggesting an involvement in disease perpetuation and/or chronicity.

Rheumatoid arthritis (RA) is a common and chronic autoimmune disease, with a prevalence of 0.5–1% in most populations, and affects 3 times more females than males. RA is a complex trait with a strong genetic component (1, 2), and several major histocompatibility complex (MHC) and non-MHC susceptibility loci have been identified in recent years (3–7). While very recent studies based on a positional cloning strategy identified the first genes implicated in susceptibility to RA (8, 9), the identities of most susceptibility and severity regulatory genes remain largely unknown. Among the difficulties involved in gene identification are the inherent characteristics of complex traits, which include variable penetrance and the potential interactions (epistasis) between disease alleles that lead to significant differences in clinical expression of disease (10). Additionally, yet-unidentified sex factors appear to regulate the penetrance of certain disease alleles, making these disease alleles sex-specific (11–14).

It has been hypothesized that the genes controlling susceptibility and development of the autoimmune processes during the initial stages of disease are different from those regulating disease severity and later stages in the disease course. Therefore, identifying genes that regulate different stages of RA progression and disease severity is likely to generate more, and perhaps better, targets for the development of more efficient therapies. However, none of the RA genetic studies based on genome-wide linkage analyses in families with multiple affected individuals have been designed to study severity, disease stages (early versus late or chronic), or outcomes. These disease characteristics can be more easily studied in rodent models of autoimmune arthritis. Specifically, disease onset can be timed with reasonable precision, and arthritis severity followed prospectively, with reduced genetic diversity and environmental noise. Promising findings have emerged from studies of experimental autoimmune erosive arthritis in rodents aimed at identifying both susceptibility and severity regulatory loci (15–22), as well as loci that regulate different stages of the disease course (15, 23).

We have focused our analyses on collagen-induced arthritis (CIA) severity, and have previously identified 3 non-MHC regulatory loci in a (DA × ACI)F2 intercross, Cia5, Cia7, and Cia10, and a major sex effect, which account for 11%, 20%, 14%, and 10% of the genetic contribution to the variance in the arthritis scores, respectively (17). Following our observations in DA.F344(QTL) (quantitative trait locus)–congenic strains demonstrating that certain loci regulate arthritis in a sex-specific manner (11–13), we reasoned that the same could occur in the sex-influenced (DA × ACI)F2 intercross, and that further analyses were warranted.

Genome-wide screens done separately in males and females identified 2 novel, recessive, female-specific arthritis severity regulatory loci, Cia25 and Cia26, and revealed that Cia26 is also a chronicity locus in epistasis with Cia7. Ncf1 was found to be a strong candidate gene for Cia25, and polymorphisms between the DA and ACI strains were identified.



Inbred specific pathogen–free DA (DA/Bkl) (CIA-susceptible) and ACI (ACI/Hsd) rats (CIA-resistant) were purchased from Bantin & Kingman (Fremont, CA) and Harlan Sprague-Dawley (Indianapolis, IN), respectively. F1 and, subsequently, F2 progeny were generated. Offspring from both (DA × ACI) and (ACI × DA) mating pairs developed similar disease severity and were used in combination for the (DA × ACI)F2 intercross analyses. All experiments involving animals were reviewed and approved by the Institutional Animal Care and Use Committee.

Induction of CIA.

Homologous rat type II collagen (RII) was prepared as previously described (17, 24). Briefly, purified collagen was dissolved in 0.1N acetic acid (1 mg/ml at 4°C) and emulsified with an equal volume of cold Freund's incomplete adjuvant (IFA). Eight-week-old rats received intradermal injections on the back at a dose of 2 mg/kg. A booster injection of the RII/IFA emulsion (100 μg of collagen in 0.1 ml) was injected at the base of the tail on day 7.

Arthritis scoring.

We used a previously described arthritis scoring system (16, 17, 25) that evaluates individual joints and weights the arthritis severity by joint size, as follows: (a) for the interphalangeal, metacarpophalangeal, and metatarsophalangeal joints, each of the 4 lateral digits was scored as 0 or 1 (0 = no arthritis and 1 = arthritis present); and (b) for the wrist, mid-forepaw, ankle, and midfoot joints, each was scored on a scale of 0–4 (0 = normal, 1 = minimal swelling, 2 = moderate swelling, 3 = severe swelling, and 4 = severe swelling and non–weight bearing). The scores from all involved joints were added (maximum score per rat 80). A single observer recorded the day of arthritis onset (following immunization) and obtained the scores twice a week during the 39-day observation period. Arthritis scores have been previously demonstrated to correlate with histologic and radiologic abnormalities. Proper immunization was confirmed by measuring antibodies to RII.

Antibodies to RII.

Blood was collected on day 28 postimmunization for serum isolation. Antibodies to RII were measured according to a previously reported enzyme-linked immunosorbent assay method (26).

Delayed-type hypersensitivity (DTH) responses to RII.

Intradermal DTH responses were used as a T cell antigen–presenting cell–dependent phenotype. DTH was measured at the end of the experiment, on day 42 postimmunization, as previously described (27). Briefly, the abdomen was shaved, and a 50-μl injection containing 50 μg of RII diluted in 0.2% NaCl, 0.05M Tris HCl, pH 7.2, was injected intradermally. The 2-direction diameter of the response (induration) was measured with calipers 72 hours after the injection, and the mean (in mm) was used for analysis.

DNA extraction and genotype analysis.

Animals were killed at the end of the observation period. DNA was isolated from the livers, and genotypes were determined by polymerase chain reaction (PCR) amplification of polymorphic DNA simple sequence-length polymorphisms (SSLPs) distinguishing DA from ACI alleles. PCR primers were either synthesized or purchased from Research Genetics (Huntsville, AL). Information about the primer sequences is available online at the National Institutes of Health (NIH) Rat Genetic Database (28) and at the Rat Genome Database (29). All PCRs were set up in 10-μl reactions. Amplification conditions have been previously described (17, 28, 30).

32P-labeled PCR products were resolved in 8% polyacrylamide sequencing gels, followed by autoradiography, while fluorescence-labeled products (using forward-labeled primers) were run in ABI 310 capillary genotypers/sequencers (ABI, Foster City, CA). GeneScan 3.1 and Genotyper 1.1 software (ABI) were used for data extraction and allele assignment of the fluorescence-labeled PCR products. All genotypes were checked by at least 2 readers, and questionable readings were rechecked, repeated, or omitted. Genetic maps were constructed using MapMaker/EXP version 3.0b (31, 32). The combined (DA × ACI)F2 male and female genetic maps have been previously reported (17) and are available at the NIH Rat Genetic Database (28).

Genome-scan strategy.

More than 3,000 markers (28, 29, 33–36) were screened in order to identify the 86 polymorphic markers used in the original genome-wide scan. To study arthritis severity, F2 rats with arthritis scores greater than zero (n = 327; 143 males and 184 females) were used for QTL analysis. Based on sex differences in maximum arthritis scores (MAS) (median in females 22 and median in males 12; P < 0.001, by Mann-Whitney rank sum test), we had previously genotyped the phenotypic extremes (based on MAS) of both F2 males and females to ensure that both groups were similarly represented in the analysis (17). Accordingly, rats with the 15% highest and 10% lowest MAS in the female group (n = 47) and the highest and lowest 15% MAS in the male group (n = 42) had been genotyped. This selective genotyping strategy has been considered to improve statistical power (37, 38).

A 2-stage strategy was used in the present study. First, male and female MAS phenotype extremes were analyzed separately in order to identify sex-specific regulatory loci. Interval QTL analysis was done with the MapMaker/QTL program version 1.1 b (37, 39, 40). Second, to search for possible epistasis in subgroup analyses, all 327 F2 arthritic rats were genotyped with the SSLP markers closest to the peaks of Cia7 (Cpb1) and Cia10 (D2Arb17), the most significant regulatory loci previously identified in the (DA × ACI)F2 intercross. All F2 arthritic rats were also genotyped with markers covering chromosomes 12 and 18; these genotyping data were used in a comprehensive QTL analysis of different phenotype parameters (number of days from immunization to disease onset, arthritis scores on 8 scoring days from day 14 to day 39 postimmunization, mean arthritis scores, cumulative arthritis scores or area under the curve, MAS, weight loss, antibodies, and DTH responses to RII) as well as to compare the mean (or median for non-normally distributed data) scores associated with the different genotypes. Additional SSLP markers were added to the genetic maps of chromosomes 12 and 18 (3 and 5 markers, respectively; see Results). Male and female rats were stratified according to the genotypes at Cpb1 and D2Arb17 to look for interacting loci.

Ncf1 sequencing.

DA/BklArb and ACI/Hsd total RNA was extracted from spleens using the Qiagen RNeasy kit (Qiagen, Chatsworth, CA), and DNA was isolated from tails using the Qiagen DNeasy kit. Complementary DNA (cDNA) (synthesized with the SuperScript II kit, Invitrogen, San Diego, CA) and genomic DNA were used for PCRs. Two pairs of primers covering the coding region containing the amino acid–changing single-nucleotide polymorphisms (SNPs) previously described by Olofsson et al (41) were used for the cDNA sequencing (Ncf1-cD-1F ACACCTTCATTCGCCACATC and Ncf1-cD-1R ATCTTTGGCCGTCAGGTATG; Ncf1-cD-2F GGCCAAAGATGGCAAGAATA and Ncf1-cD-2R GATAGGTGTCCTGGCTGAGG) (based on GenBank accession no. NM_053734.1). Two additional pairs were designed to confirm the SNPs in genomic DNA (Ncf1-gD-106F CCTT TCTGGACTTGACTTTG and Ncf1-gD-106R GCTACTCACT GGCTGTCATT; Ncf1-gD-153F ATCGTGAGCTTGACT GATCT and Ncf1-gD-153R CTCGCTTTTCTCTACGACAT) (based on GenBank accession no. NW_042785). Sequencing reactions used BigDye Terminator reagents (ABI) and were run in an ABI 3100 16-capillary sequencer. The DNAStar sequence analysis package (DNAStar, Madison, WI) was used for sequencing analyses.

Statistical analyses.

Means were compared with the t-test or one-way analysis of variance (ANOVA), and medians were compared with the Mann-Whitney rank sum test or ANOVA on ranks, using either SigmaStat 3.0 (SPSS, Chicago, IL), or Splus 3.4 (Mathsoft, Cambridge, MA).


Identification of a novel female-specific regulatory locus on rat chromosome 12 by QTL analysis of rats stratified according to sex.

The original genome-wide screen in the (DA × ACI)F2 intercross (17) identified 3 regulatory loci: 2 on chromosome 2, and 1 on chromosome 10 (Table 1). Following our previous observations in DA.F344(QTL)-congenic strains demonstrating that certain loci originally identified in the F2 intercross operate in a sex-specific manner (11–13), we sought in this study similar sex–QTL interactions in the (DA × ACI)F2 intercross rats. QTL analyses according to sex, using the phenotypic extremes of the MAS, identified a novel female-specific locus on rat chromosome 12, Cia25, with a maximum logarithm of odds (LOD) score of 3.0 (LOD score in males 1.1) and with its peak closest to D12Wox12 (Table 1).

Table 1. Identification of a new chromosome 12 locus, Cia25, by QTL analysis in (DA × ACI)F2 rats stratified according to sex*
LocusChromosomeLength (cM)Peak markerLOD scores
All (n = 327)Males (n = 143)Females (n = 184)
  • *

    Phenotype extremes of the maximum arthritis scores achieved during the observation period used in the analyses.

  • Quantitative trait locus (QTL) length based on the 2–logarithm of odds (2-LOD) score support interval.

  • Results were reported previously (17).


We considered that Cia25 could have its maximum effect at a specific stage during disease development and/or progression. To analyze the effect of this locus on the latency to disease onset (number of days from immunization to disease onset), arthritis scores at different time points (from day 14 to day 39), cumulative and mean arthritis scores, DTH, and antibody responses to RII, all 327 arthritic rats were genotyped with markers covering chromosome 12 and analyzed with MapMaker/QTL.

The strongest regulatory effect of Cia25 was observed in the arthritis scores for female rats on day 28, with a peak LOD score of 4.7 (Figure 1). High LOD scores were also observed for arthritis scores in female rats on days 25 (LOD 2.6), 32 (LOD 3.7), and 35 (LOD 2.9), as well as for the mean (LOD 3.0) and cumulative arthritis scores (LOD 3.0). LOD scores for MAS in females reached 3.7. This locus accounted for 23.4% of the genetic contribution to the variance in the arthritis scores on day 28, with a recessive or additive mode of inheritance, per MapMaker/QTL analysis.

Figure 1.

Logarithm of odds (LOD) score plots, identifying Cia25 on female chromosome 12. The phenotype analyzed is the arthritis score on day 28 following immunization. Dotted line represents level for significant evidence of linkage (58). Shaded bar at top identifies the 2-LOD score support interval. Arrow marks the position of Ncf1. Shaded triangles identify simple sequence-length polymorphism marker positions.

Arthritis scores in female rats on day 28 were further analyzed according to genotypes at D12Wox12, the marker closest to the peak of Cia25. The DD genotype (DA homozygous) was associated with significantly higher arthritis scores, compared with the AA genotype and with heterozygous (HET) rats (median score for DD, 38.5; for HET, 21; and for AA, 16) (P = 0.0004 for DD versus HET and P = 0.00003 for DD versus AA) (Figure 2). The scores for the AA and HET rats were not significantly different, a finding that supports a recessive mode of action. The DD genotype at D12Wox12 was associated with a 45% higher median arthritis score compared with the AA and HET genotypes, which underscores the importance of Cia25 in the regulation of CIA severity. Genotypes at D12Wox12 did not significantly influence arthritis scores in male rats on day 28 (Figure 3).

Figure 2.

Arthritis scores in female rats on day 28 according to genotypes at D12Wox12, the marker closest to the peak of Cia25. Results are shown as box plots. The lines inside the boxes indicate the median, the outer borders of the boxes indicate 25% and 75%, and the bars extending from the boxes indicate 5% and 95%. Circles indicate outliers. P values are 2-group comparisons by the Mann-Whitney rank sum test. P = 0.000041 by analysis of variance on ranks comparing the 3 groups. HET = heterozygote.

Figure 3.

Arthritis scores in male rats on day 28 according to genotypes at D12Wox12, the marker closest to the peak of Cia25. Results are shown as box plots. The lines inside the boxes indicate the median, the outer borders of the boxes indicate 25% and 75%, and the bars extending from the boxes indicate 5% and 95%. Circles indicate outliers. P = 0.607 by analysis of variance on ranks comparing the 3 groups. HET = heterozygous.

To our knowledge, this is the first identification of a sex-specific arthritis QTL in this region of chromosome 12. None of the other phenotypes we tested were linked to chromosome 12.

Additional loci identified in the genome-wide scan and their sex effect in the (DA × ACI)F2 intercross.

Table 1 shows the LOD scores for male and female rats at the 3 loci previously reported in the combined analyses. Specifically, LOD scores at Cia7 were similar in both males and females (peak LOD scores of 3.1 and 2.7, respectively, closest to Cpb1). This locus operates in a dominant manner. The sex-specific analysis reduced the sample size and, thus, reduced the LOD score from the original combined analyses, which was 4.6.

The peak LOD score at the other chromosome 2 locus, Cia10, reached 3.0 in the male group, but only 1.3 in the female group (Table 1). These results indicate that Cia10 is a male-specific locus, and that the male effect accounted for most of the LOD score of 3.4 that was identified in the combined analysis of males and females. No other sex-specific loci were identified, and the LOD scores for Cia5 were decreased both in males and in females due to the sample size reduction.

Identification of a new Cia7-dependent chromosome 18 locus by analysis according to a 2-locus epistatic model.

A significant percentage of the genetic contribution to the arthritic phenotype remained unaccounted for after the previously reported genome-wide screen. We considered that additional genetic loci regulating arthritis severity could be in epistasis with either Cia7 or Cia10 and that their identification would require the inclusion of genotype data at the latter loci as variables in the analyses. MapMaker/QTL analysis had previously suggested a dominant mode of action for Cia7, which was supported by comparing the MAS according to the genotypes at Cpb1, the marker closest to the QTL peak (median score for AA, 16; for HET, 24; for DD 26) (P < 0.001 for DD versus AA, P = 0.002 for HET versus AA, and P not significant for DD versus HET). Therefore, rats with genotypes DD and HET were grouped together and analyzed separately from rats with the AA genotype group.

A similar procedure was undertaken to look for interactions with Cia10. While MapMaker/QTL could not differentiate between a recessive or additive mode of inheritance for Cia10, comparison of the MAS according to the 3 possible genotypes at D2Arb17, the marker closest to the QTL peak, confirmed a recessive mode (median MAS for AA, 18; for HET, 21; for DD, 33) (P < 0.05 for DD versus AA and versus HET; P not significant for AA versus HET). Therefore, rats that were AA and HET at D2Arb17 were grouped together and analyzed separately from DD rats for interactions with other loci. Males and females were analyzed separately.

In addition to chromosomes 2 and 12, all rats were genotyped with markers covering chromosome 18 for the purpose of identifying epistatic interactions. The focus on chromosome 18 was based on 3 previous reports that suggested that it might contain an autoimmune arthritis locus. First, a study of experimental allergic encephalomyelitis using the same inbred strains studied herein ([DA × ACI]F2 intercross) identified a disease locus on chromosome 18 (42), raising the possibility that the same locus could regulate additional forms of autoimmune disease, as previously suggested (43, 44). Second, this chromosome contains a major sex-specific rat autoimmune arthritis (CIA)–regulatory locus identified in a (BB × BN)F2 intercross study (45). Third, 2 independent groups of investigators have recently identified an RA susceptibility locus in the syntenic region of human chromosome 18 (3, 5). These studies strongly suggested that rat chromosome 18 could contain an autoimmune arthritis–regulatory gene in the (DA × ACI)F2 intercross, and since this locus had not been identified in the original genome-wide screen, we considered that it could be in epistasis with either Cia7 or Cia10.

QTL analyses of chromosome 18 in female rats carrying at least 1 D allele (DD and HET groups) at Cpb1/Cia7 identified a new female-specific recessive locus, Cia26, located between markers D18Mit9 and D18Rat60 (Figure 4), with a peak LOD score of 3.6 for arthritis scores on day 39 postimmunization. Cia26 accounts for 30% of the genetic contribution to the variance in arthritis scores in this female subgroup, and its 2-LOD score support interval extends for 16 cM. The Cia26 locus effect on arthritis scores could be observed from day 32 postimmunization onward (peak LOD scores of 2.5 for arthritis scores on day 32 and 2.97 on day 35), but not in prior arthritis scores or in the MAS. Arthritis scores on day 39 were significantly higher (37% increase) in rats with the DD genotype at D18Mit9 (marker closest to the peak of Cia26) compared with rats with the HET or AA genotype (mean ± SEM for DD, 20.7 ± 2; for HET, 13.9 ± 1.3; for AA, 11 ± 1.7) (P = 0.003 for DD versus HET, P = 0.001 for DD versus AA, and P = 0.26 for HET versus AA), supporting a recessive mode of action. Similar analyses were done in the female group of rats with the AA genotype at Cpb1/Cia7; however, the LOD scores were not significant. No evidence for a chromosome 18 locus was identified in males (Figure 4).

Figure 4.

Logarithm of odds (LOD) score plots, identifying Cia26 on female chromosome 18. The phenotype analyzed is the arthritis score on day 39 following immunization in rats carrying genotypes DD or HET (heterozygous) at Cpb1/Cia7. Dotted line represents the level suggestive of linkage (58). Shaded bar at top identifies the 2-LOD score support interval. Shaded triangles identify simple sequence-length polymorphism marker positions.

The epistatic interaction between genotypes at Cpb1/Cia7 and D18Mit9/Cia26 was also analyzed using one-way ANOVA (Figure 5). Based on the Cia26 recessive mode of action, female rats with the AA or HET genotype at D18Mit9 were grouped together to compare their arthritis scores on day 39 with those of female rats with the DD genotype. As demonstrated in Figure 5, female rats with at least 1 D allele at Cpb1/Cia7 (DD or HET) and 2 D alleles (DD genotype) at D18Mit9/Cia26 had a significantly higher mean arthritis score compared with all other groups. In the presence of the AA genotype at Cpb1/Cia7, the arthritis scores were significantly reduced, independently of the genotypes at D18Mit9/Cia26, demonstrating epistasis as defined by Bateson (10, 46). No interactions between genotypes at Cia7 and Cia26 were identified in the males (Figure 6). Taken together, the data suggest that Cia26 regulates the processes involved in disease perpetuation and chronicity that take place at later stages in the disease course in females, and operates in a Cia7-dependent epistatic manner.

Figure 5.

Arthritis scores of female rats on day 39 according to genotypes at Cpb1/Cia7 and D18Mit9, the marker closest to the peak of Cia26, demonstrating the interaction between the 2 loci in arthritis scores. Values are the mean and SEM. P values were determined by one-way analysis of variance. HET = heterozygous; N = number of rats in the group.

Figure 6.

Arthritis scores of male rats on day 39 according to genotypes at Cpb1/Cia7 and D18Mit9, the marker closest to the peak of Cia26. Values are the mean and SEM. No significant interactions were seen between these 2 loci (P = 0.160 by one-way analysis of variance). HET = heterozygous; N = number of rats in the group.

No interactions were found between Cia10 and Cia26 and/or Cia25. LOD scores at Cia25 were not influenced by the genotype at Cpb1/Cia7. None of the other phenotypes studied were linked to chromosome 18.

Polymorphisms on Ncf1, a candidate gene for Cia25.

The same arthritis-susceptibility and arthritis-resistance SNPs described by Olofsson et al (41) as correlating with variable production of reactive oxygen species were detected in the DA/BklArb and ACI/Hsd strains (Table 2). Specifically, DA/BklArb rats had an A allele, whereas ACI/Hsd rats had a G at SNP nucleotide 316 (codon 106). DA/BklArb rats had a T allele, whereas ACI/Hsd rats had a C at SNP nucleotide 458 (codon 153). Both SNPs cause amino acid changes in their respective codons.

Table 2. Ncf1 polymorphisms in DA and ACI rats*
 SNP at nt 316AA on codon 106SNP at nt 458AA on codon 153
  • *

    SNP = single-nucleotide polymorphism; nt = nucleotide in the cDNA sequence; AA = amino acid.



Rat models of autoimmune arthritis studied in intercrosses generated from inbred strains that differ in their susceptibility to and severity of disease provide a powerful tool for facilitating the identification of loci that regulate different stages in the disease course (15, 23), as well as specific disease characteristics, such as disease severity (16, 17, 25, 47) and the production of rheumatoid factors (48) and acute-phase reactants (49). The reduced genetic diversity and relative ease in generating large numbers of offspring as well as in doing selective breeding facilitate not only the identification of the genetic loci but also the study of potential epistatic interactions, as described herein. Additionally, the phenotypic effect of loci identified in linkage studies can be confirmed in congenic strains (11–13, 22, 50), and those studies will ultimately lead to gene identification (41). Findings from studies in rats can then be brought into studies of humans with RA for more focused analyses (51).

Clinical disease onset in rat CIA typically occurs around day 14 postimmunization, with maximum severity observed between days 21 and 25, followed by a progressive and slow improvement (13, 52). This clinical improvement is followed by a chronic stage, considered to occur from day 35 to day 60 (53). Therefore, studying arthritis severity at different time points provides an opportunity to identify genes/QTL involved in processes specific to that stage, such as disease initiation, MAS, and chronicity.

To facilitate the identification of the non-MHC loci that regulate disease severity, we previously generated an intercross between the DA/Bkl and ACI/Hsd rat strains, which share the same arthritis-prone MHC RT1av1 haplotype, yet differ in their susceptibility to CIA. A previous genome-wide scan in the (DA × ACI)F2 intercross identified a major sex effect (not linked to chromosome X) that accounted for 10% of the variance in arthritis severity. Two novel autosomal loci on chromosome 2 (Cia7 and Cia10) and a replication of a locus on chromosome 10 (Cia5) have been reported (17). These loci and the sex effect accounted for 55% of the genetic contribution to the variance in MAS, suggesting the existence of yet-unidentified additional loci.

Our subsequent studies with DA.F344(QTL)-congenic strains revealed that certain loci operate in a sex-specific manner, regulating disease in one sex but not in the other (11–13). These observations led us to hypothesize that additional loci existed in the original sex-influenced (DA × ACI)F2 intercross and operated in a sex-specific manner. In fact, analyses of the F2 population according to sex identified a novel female-specific recessive locus on chromosome 12, Cia25, that had not been detected in the study of females and males combined. Cia25 had its maximum effect on the arthritis scores on day 28 postinduction (LOD 4.7). Cia25 also influenced arthritis scores from day 25 to day 35, the MAS, and the cumulative scores, suggesting that this locus may regulate not only disease severity, but also processes involved in disease perpetuation, duration, and/or chronicity. Cia25 colocalizes with Cia12 (54) and Pia4 (15, 41), loci previously found to regulate CIA in (DA × BN)F2 rats and pristane-induced arthritis (PIA) in (DA × E3)F2 rats (Figure 7). Interestingly, Cia12 also exhibits sex specificity (Griffiths MM, et al: unpublished observations). Additionally, the mouse chromosome 5 syntenic region also contains autoimmune arthritis-susceptibility and arthritis-severity loci mCia11 and Pgia16 (55), further supporting the relevance of this locus (Figure 7).

Figure 7.

Homology map showing Cia25 on rat chromosome (chrom.) 12, mouse chromosome 5, and the human cytogenetic syntenic bands, as well as the autoimmune trait regulatory loci located in those regions. RH = radiation hybrid (map); Cia = collagen-induced arthritis; Eae = experimental allergic encephalomyelitis; Eau = experimental allergic uveitis; Pia = pristane-induced arthritis; Pgia = proteoglycan-induced arthritis; mCia = murine CIA; Dssc = dextran sulfate sodium–induced colitis; Bb = Borrelia burgdorferi; Lbw = NZB/NZW murine lupus locus; Ear = exogenous antigen response; SLE = systemic lupus erythematosus; UC = ulcerative colitis; IBD = inflammatory bowel disease; MS = multiple sclerosis. (Maps constructed according to and

Olofsson et al (41) recently positionally identified the gene that accounts for Pia4 as Ncf1, a NADPH oxidase complex–related gene. Ncf1 is located within Cia25 (Figure 1) and, therefore, is a potential candidate gene. While Pia4/Ncf1 has its maximum effect around the same period in the disease course as Cia25, Pia4/Ncf1 does not operate in a sex-specific manner and has an additive mode of action, while Cia25 is recessive and female-specific; therefore, it remains unclear whether Cia25 and Pia4 are accounted for by the same gene. However, we have sequenced DA/Bkl and ACI/Hsd cDNA and genomic DNA and were able to identify the same Pia4 disease-susceptibility (DA) and disease-resistance (E3) polymorphisms described by Olofsson et al (41). The presence of the same alleles suggests that Ncf1 accounts for Cia25, or at least part of its effect, in the (DA × ACI)F2 intercross. Congenic strains, currently at their final stages of breeding, will be critical to determining the specific gene(s) that accounts for Cia25.

The sex-specific genome-wide screen also revealed that the male group accounted for the Cia10 effect described in the original study (17). Cia10 and its syntenic regions in human and mouse chromosomes are of great interest because they contain loci that regulate inflammation (56), autoimmune diabetes (Iddm3) (57), and several other autoimmune traits (for review, see ref. 17), suggesting that this gene could also be involved in the regulation of other autoimmune diseases.

The identification of Cia26 on chromosome 18 and its epistatic interaction with Cia7 underscore the difficulties involved in identifying complex-trait disease genes. In the presence of the AA genotype at Cpb1/Cia7, the arthritis scores were significantly reduced independently of the genotypes at D18Mit9/Cia26; this demonstrates a suppressive effect of the first locus on the second, defining epistasis as described by Bateson (10, 46) (Figure 5). Cia26 is a recessive locus and influences the arthritic phenotype from day 32 to 39, suggesting that this gene is involved in the regulation of disease perpetuation, duration, or the more chronic stages of disease. Despite the 2-stage subset analyses required for the identification of Cia26 (sex and Cpb1/Cia7), the peak LOD score still reached 3.6. This Cia26 LOD score was even higher than that found for Cia7 in a comparable female subgroup (LOD 2.7) and higher than the stringent LOD 2.8 criteria recommended for “suggestive for linkage” (58), indicating that Cia26 has a significant contribution to the phenotype. Additionally, since Cia26 appears to regulate disease chronicity, and the LOD scores progressively increased from day 32 to day 39, it is conceivable that a longer period of observation might identify the time point of maximum effect of this locus.

Cia26 is also of great interest because it colocalizes with Cia17 and Pia15, loci that regulate CIA and PIA in the (BB × BN)F2 and (DA × E3) × DA BC1 strains, respectively (45, 59). Cia17, but not Pia15, also operates in a female-specific manner. Furthermore, the mouse and human syntenic regions (both on chromosome 18) also harbor loci that regulate murine CIA and proteoglycan-induced arthritis (mCia18 and Pgia11) (55) and, more importantly, RA (3, 5), supporting the concept that 1 or more real arthritis-regulatory genes are located in this region.

Similar to Cia5, Cia7, and Cia10, both Cia25 and Cia26 are in chromosomal regions that also contain loci that regulate different forms of autoimmune diseases in rats (42, 60–62) and are syntenic to regions that regulate autoimmune diseases or traits in mice (63–68) and humans (44, 69–76) (Figure 7). The accumulation of autoimmunity-regulatory loci to these regions certainly raises the previous suggestion that some of those genes, including Cia25 and Cia26, will be common to different forms of autoimmune diseases (43, 44).

Sex has only recently been considered a covariate in genetic studies in RA (3, 77), and more comprehensive analyses remain to be done.

In conclusion, we have identified 2 novel female-specific CIA loci in the (DA × ACI)F2 intercross that regulate disease severity and disease persistence, duration, and/or chronicity. One of these loci, Cia26, is in epistasis with Cia7. We have provided evidence suggesting that Ncf1 may account for Cia25. The definitive characterization of the phenotypic effect of these 2 loci and the others identified in this intercross, as well as the precise gene identification, will require further breeding of congenic strains. We consider that genetic analyses and congenic studies in rodents have the potential to facilitate the identification of epistatic loci for future and focused studies in humans.