Sex effect on clinical and immunologic quantitative trait loci in a murine model of rheumatoid arthritis

Authors


Abstract

Objective

To explore the effect of sex on clinical and immunologic traits in major histocompatibility complex–matched (H-2d) F2 hybrid mice with proteoglycan (PG)–induced arthritis and to identify how the quantitative trait locus (QTL) on the X chromosome influences the onset QTL of another chromosome.

Methods

(BALB/c × DBA/2)F2 hybrid mice were immunized with cartilage PG, and a genome-wide linkage analysis was performed using >200 simple sequence-length polymorphic markers. The major clinical traits (susceptibility, onset, and severity) were assessed, and PG-specific T and B cell responses, and the production of proinflammatory and antiinflammatory cytokines (tumor necrosis factor α, interleukin-1 [IL-1], IL-6, interferon-γ, IL-4, IL-10, and IL-12) were measured in 133 arthritic and 426 nonarthritic female and male F2 hybrid mice. The major clinical and immunologic traits were linked to genetic loci, and potential linkages among these QTLs and the effect of sex were analyzed.

Results

Thirteen QTLs reported in previous studies were confirmed. Binary traits (susceptibility to arthritis) and disease onset were female specific and were identified on chromosomes 3, 7, 10, 11, 13, and X. QTLs for disease severity were mostly male specific and were located on chromosomes 1, 4, 5, 8, 14, 15, and 19. In addition, we identified 4 new QTLs for the onset of arthritis on chromosomes 3, 4, and 11, and 1 new QTL for severity on chromosome 14; all showed a strong gender association. A locus on the X chromosome interacted with a QTL on chromosome 10, and these 2 loci together seemed to control disease incidence and onset. Most of the clinical traits (QTLs) shared common regions with the immunologic traits and frequently showed a locus–locus interaction.

Conclusion

Numerous immunologic QTLs overlap with clinical QTLs, thus providing information about possible mechanisms underlying QTL function. Disease susceptibility and onset showed predominant linkage with the female sex, under the control of a QTL on the X chromosome, while the severity QTLs were more strongly linked to the male sex.

Rheumatoid arthritis (RA) is a complex inflammatory autoimmune disease with a female prevalence that affects ∼1% of the population (1–4). In spite of extensive epidemiologic, genetic, and immunologic studies, the etiology of RA remains unknown. Because of the difficulty of examining early events in disease progression, most studies concentrate on inflammation in peripheral joints when and where the clinical symptoms become evident. Linkage, or quantitative trait loci (QTL), analysis is a unique method of identifying susceptibility genes, at least theoretically, irrespective of the time point at which the gene was, or is still, active. Linkage analysis of families with RA probands is in progress, and several chromosomal regions (QTLs) have already been identified (5–14). Among the disease-associated loci, the major histocompatibility complex (MHC) on human chromosome 6 has the strongest effect on RA susceptibility, as it does in many other autoimmune diseases (5, 6, 11, 12, 14, 15).

The importance of sex in the pathogenesis of RA has long been recognized, and the female prevalence of RA is well-documented (1, 16, 17). The relationship between arthritis and sex in different animal models of RA is not evident, although many investigators have reported significant sex-related differences in arthritis severity and incidence in inbred murine strains, their F2 hybrids, and in congenic strains (18–23). The X and Y chromosomes are therefore logical targets of linkage analysis when the sex effect of disease-associated traits is at the center of a genetic study. Only a very few linkage studies, however, have confirmed the involvement of the X chromosome in humans with RA (6, 14, 15) or in experimental models of arthritis (18, 20, 23).

QTL analysis in humans is hampered by the genetic heterogeneity of the population, the uncontrollable environmental effects, and the limited number of meioses. Therefore, animal models of RA are attractive tools because use of such models not only overcomes genetic complications, but also permits studies during the early phases of disease development. Arthritis can be induced in susceptible rodent strains by injection of adjuvant (24), oil (25), collagen (26), bacterial cell wall components (27), or cartilage proteoglycan (PG) (28, 29). To date, linkage analysis studies in animal models of RA have identified more than 20 loci that are either specific to the given animal model or shared with other models or with RA in humans. A number of traits of arthritis, including the severity, onset, and susceptibility and the production of antibodies and cytokines, have been linked to chromosome regions that are syntenic in various species and often show overlapping localization in several autoimmune diseases, thus suggesting shared genetic components (27, 30–32).

Cartilage PG (aggrecan)–induced arthritis (PGIA) can be induced in genetically susceptible BALB/c or C3H/HeJCr mice (28, 29, 33–35). PGIA shows many similarities to RA, as indicated by the clinical symptoms and the findings of radiographic, laboratory, and histopathologic examination of peripheral joints (28, 29, 33, 36, 37). In a special intercross, using F2 hybrids of DBA/1 mice, which are susceptible to collagen-induced arthritis (CIA) but not to PGIA, and PGIA-susceptible mice, which are resistant to CIA, we were able to compare the effect of the MHC in 2 different autoimmune arthritis models with shared genetic backgrounds (37, 38). Based on the results of these studies, we concluded that the MHC locus plays a critical role in PGIA, although the effect is not as strong as that in CIA. To date, we and other investigators have identified 18 Cia and 25 Pgia loci in mice (34, 35, 37–43); 14 of these QTLs were colocalized in the 2 models, and some of these loci corresponded to QTLs identified in humans with RA.

These genetic studies suggested that while numerous QTLs control and/or modify the clinical and immunologic traits, the MHC plays a predominant role. To exclude the effect of the MHC on some QTLs and to detect possible interactions among other QTLs, we used MHC-matched (BALB/c × DBA/2)F2 hybrid mice. This particular genetic cross also allowed us to study the effects of sex on the clinical and immunopathologic features of the disease and on QTLs. In the present study, we used a population of F2 hybrids (n = 559), separated the arthritis trait into susceptibility and severity subtraits, and compared the immunologic/pathophysiologic markers in all immunized animals.

MATERIALS AND METHODS

Animals, antigen, and immunization.

Female BALB/c mice and male DBA/2 mice (Charles River, Kingston, NY) were mated, and the F1 offspring were intercrossed to generate F2 hybrids (n = 559). At the age of 12 weeks, mice were immunized with cartilage PG using standard immunization protocols (33). Briefly, 100 μg of human cartilage PG (measured as protein) was emulsified in adjuvant (100 μl) and injected intraperitoneally on days 0, 7, 28, and 49. Freund's complete adjuvant (Difco, Detroit, MI) was used for the first and fourth injections; the second and third injections contained antigen in Freund's incomplete adjuvant. Mice were killed 7–8 weeks after the fourth injection (i.e., on days 102–104 of the experiment).

Assessment of arthritis and clinical traits.

Arthritis was assessed 3 times each week. Inflammation in each paw was scored on a 0–4 scale, and the findings in the 4 paws were summed to yield an arthritis score (range 0–16 for each animal) (28, 33, 34, 36). Paws with questionable clinical scores and those with an arthritis score of 1 were evaluated histopathologically, as described previously (33, 35). The clinical score includes both qualitative and quantitative traits, such as the incidence of arthritis (susceptibility) and the severity of inflammation, but it does not reflect the time of arthritis onset (i.e., how early the arthritis develops after immunization).

The primary clinical score in an autoimmune model is the binary (qualitative) trait, which is the susceptibility to disease. This trait has only 2 values: 0 for nonarthritic animals and 1 for arthritic animals. All other components of the clinical score for arthritis are quantitative: disease severity, disease progression, and time to disease onset. Therefore, we separated the qualitative (susceptibility) and quantitative (other clinical) traits, and introduced additional scores based on the arthritis index. The severity score is the same as the arthritis score, but it applies only to arthritic mice (range of scores 1–16). In addition, a disease onset score (range 0–5) that reflected how quickly the animals developed arthritis was assigned. A maximum score of 5 was given for all animals that developed PGIA on day 50 or earlier. An onset score of 0 was given for animals that never showed any symptom of inflammation and did not develop arthritis by the end of the experimental period (days 102–104). Intermediate scores between 0 and 5 were assigned using linear time adjustments.

Measurement of antibodies, T cell responses, and cytokine production.

Antibodies to the immunizing (human) and mouse (self) cartilage PGs were determined by enzyme-linked immunosorbent assay (ELISA) as described elsewhere (38, 44). Briefly, 96-well Maxisorp plates (Nunc International, Hanover Park, IL) were coated with either chondroitinase ABC–digested human (for heteroantibodies) or native mouse (for autoantibodies) cartilage PGs (0.1 μg of antigen protein/well). PG-specific antibodies were measured in serially diluted (1:500–1:62,500) immune sera using peroxidase-conjugated goat anti-mouse IgG, IgA, and IgM (for total antibodies) or anti-IgG1 or anti-IgG2a (for Th2- and Th1-supported IgG isotypes, respectively) secondary antibodies (Zymed, South San Francisco, CA). Serum antibody levels were expressed in arbitrary units, which were calculated in each case as a ratio of the serum dilution of the experimental sample relative to the dilution of the standard (pooled arthritic BALB/c serum) at the median of the maximum and minimum absorbance levels measured on the same plate (35).

Antigen-specific T cell responses were measured in quadruplicate samples of spleen cells (3 × 105 cells/well) cultured in the presence of 100 μg of PG protein/ml. Interleukin-2 (IL-2) was measured in supernatants collected on day 2, by determining the proliferation rate of the IL-2–dependent CTLL-2 cell line. Antigen-specific T cell proliferation was assessed on day 5, by determining the incorporation of 3H-thymidine (45, 46). In both cases, the antigen-specific response was expressed as a stimulation index, which is the ratio of the counts per minute of 3H-thymidine incorporated in antigen-stimulated cultures relative to the cpm incorporated in unstimulated cultures (33, 45). Antigen-specific production of interferon-γ (IFNγ) and IL-4 by T cells was determined in identical culture conditions as described for T cell proliferation in 4-day conditioned media (2.5 × 106 mononuclear cells/ml) using capture ELISAs (R&D Systems, Minneapolis, MN).

Serum IL-1 concentrations were determined by bioassay using D10S cells as described (35, 47). Soluble CD44 (sCD44; used as a marker of inflammation) levels in serum were determined by an ELISA developed in our laboratory (48). Serum levels of tumor necrosis factor α (TNFα), IL-6, IL-10, and IL-12 were assayed using capture ELISAs (R&D Systems or PharMingen, San Diego, CA).

Genome screening and linkage analysis.

Genomic DNA was isolated from (BALB/c × DBA/2)F2 hybrid mice and subjected to genotyping. Simple sequence-length polymorphic (SSLP) markers (MWG Biotech, High Point, NC) were used for polymerase chain reaction, which was followed by gel electrophoresis in 3.5% MetaPhore agarose (FMC Bioproducts, Rockland, ME) as described previously (34, 35, 37). These markers covered all 19 autosomes and the X chromosome; the Y chromosome was not analyzed in this study.

We initially prescreened an approximately equal number of arthritic (n = 133) and nonarthritic (n = 143) mice using a set of markers that covered all chromosomes. Along with this initial prescreening process, at least 3 markers were used for each chromosome, and 4–5 markers were used for larger chromosomes. The genomic screening was subsequently extended by increasing the number of primers and by screening all 559 PG-immunized F2 mice. When a QTL with a logarithm of odds (LOD) score ≥2.0 was found to be linked with any arthritis (clinical) score, the chromosomal region was saturated with additional markers, and linkage analysis was performed for the entire F2 population. This process was repeated several times to reach a reasonable density of markers with an average distance of <10 cM.

The main source of data for SSLP polymorphisms between BALB/c and DBA/2 progenitor strains was the Jackson Laboratory Web site (http://www.informatics.jax.org). Linkage map construction and traits–markers linkage/regression analyses were performed using Map Manager QTX version 13 software (49). The LOD threshold for suggestive linkage was set at 2.8 and at 4.3 for significant linkage (50). A permutation test to establish empirical LOD score thresholds was used (49, 51). The order of markers and their exact positions on chromosomes were confirmed using genomic maps from the Celera Discovery System (52).

Statistical analysis.

Statistical analysis was performed using SPSS statistical software (version 10.0.5; SPSS, Chicago, IL). Since some clinical traits demonstrated nonparametric distribution in the F2 hybrid population, we used the Mann-Whitney U test to examine differences between populations and Spearman's correlation to evaluate biases between traits. Chi-square statistics were used to determine the significance of locus–locus interactions, and Student's 2-sample t-test was used to compare the results from 2 groups, where the data showed normal distribution. P values less than 0.05 were considered significant.

RESULTS

Influence of sex on major clinical traits of arthritis in MHC-matched (BALB/c × DBA/2)F2 hybrid population.

As expected, based on the results of previous studies (28, 29, 34, 53), 100% of the parental BALB/c female mice developed PGIA, and the F1 hybrid mice were completely resistant. Approximately 24% of the (BALB/c × DBA/2)F2 population (133 of the 559 immunized mice) developed the disease, with a similar incidence in females (22.5%) and males (25.3%) (Figure 1).

Figure 1.

Distribution of arthritis onset scores in parental BALB/c female mice and DBA/2 male mice and their F1 and F2 hybrid offspring. Each shaded circle represents an arthritic animal; numbers within boxes represent the proteoglycan-immunized nonarthritic animals (onset score 0). See Materials and Methods for an explanation of the scoring system. Horizontal bars show the mean scores for the F2 female and male hybrids; values beside the bars show the mean ± SEM onset scores ( = P < 0.05).

The severity of arthritis was slightly higher in males than in females, but the difference was not statistically significant (data not shown). Since sex differences in disease onset have been reported in the parental BALB/c strain (29), we also analyzed the onset score in the (BALB/c × DBA/2)F2 population. We found that F2 hybrid males developed arthritis sooner (mean ± SEM onset score 2.32 ± 0.25, which corresponds to a mean day of onset 77.5) than did F2 hybrid females (onset score 1.40 ± 0.23, which corresponds to a mean day of onset 86.5) (Figure 1).

Effect of sex on immune responses and cytokine production in PGIA.

We found significant statistical differences in disease onset, with males developing arthritis an average of 9 days sooner than females (Figure 1), but no sex-related differences in either severity or susceptibility were evident in the entire F2 hybrid population. To find an explanation for this, we analyzed the 302 females and 257 males separately for T cell and B cell responses and for the production of proinflammatory and antiinflammatory cytokines.

Comparison of arthritic and nonarthritic (either female or male) F2 mice yielded no significant differences in antigen-specific T cell responses (measured as T cell proliferation or IL-2 production) or in the amounts of antigen-specific antibodies or heteroantibodies in serum (Table 1). This was not the case, however, when Th1- and Th2-supported immunoglobulin isotypes and antigen-specific production of IFNγ and IL-4 were analyzed and compared (Table 1). Thus, while no significant differences were found when the overall immune responses (either T cell– or B cell–mediated) were compared in immunized F2 hybrid mice, more sensitive markers revealed significant differences in PG-immunized arthritic versus nonarthritic F2 populations (Table 1 and Figure 2).

Table 1. Comparison of arthritic versus nonarthritic mice and of male versus female (BALB/c × DBA/2)F2 mice immunized with cartilage PG*
TraitArthritic (n = 133)Nonarthritic (n = 426)PFemale (n = 302)Male (n = 257)P
  • *

    Values are the mean ± SEM. Two-sample t-test, assuming unequal variances, was applied for comparison of means; significance was set at P < 0.05. Mann-Whitney U tests gave similar results. See Materials and Methods for a detailed description of the pathophysiologic traits. PG = proteoglycan; NS = not significant; IL-2 = interleukin-2; CM = conditioned medium; IFNγ = interferon-γ; TNFα = tumor necrosis factor α.

  • Bioassay of the IL-2–dependent CTLL-2 cell line was used to measure antigen (PG)–specific T cell stimulation, and direct 3H-thymidine incorporation was used to measure T cell proliferation.

  • Levels of serum autoantibodies and heteroantibodies to mouse and human PGs were measured by enzyme-linked immunosorbent assay. Ratios of IgG1 to IgG2 antibodies against human or mouse cartilage PG are shown. Antibody levels are expressed as arbitrary units, calculated by comparing serum levels with levels in a standard (pooled sera from arthritic animals), as described in Materials and Methods.

  • §

    Enzyme-linked immunosorbent assays were used to measure IFNγ and IL-4 in conditioned media from antigen-stimulated T cells (2.5 × 106 cells/ml), and IL-1, TNFγ, IL-6, IL-10, and IL-12 in serum.

T cell response, stimulation index      
 Proliferation1.98 ± 0.082.15 ± 0.08NS2.13 ± 0.052.15 ± 0.13NS
 IL-21.58 ± 0.101.70 ± 0.04NS1.72 ± 0.061.63 ± 0.05NS
PG-specific antibodies in serum, arbitrary units or ratio      
 Autoantibody0.75 ± 0.191.62 ± 0.220.0361.43 ± 0.241.50 ± 0.28NS
 Heteroantibody4.00 ± 0.406.08 ± 1.12NS6.33 ± 1.585.06 ± 0.59NS
 Mouse IgG1:IgG2a6.10 ± 0.307.96 ± 0.410.0146.73 ± 0.418.44 ± 0.500.008
 Human IgG1:IgG2a8.48 ± 0.5410.8 ± 0.640.0078.31 ± 0.6311.96 ± 0.600.001
 Human IgG1:mouse IgG11.96 ± 0.052.02 ± 0.04NS1.85 ± 0.052.19 ± 0.040.001
 Human IgG2a:mouse IgG2a1.59 ± 0.061.61 ± 0.04NS1.61 ± 0.041.62 ± 0.06NS
Cytokines in CM (IFNγ and IL-4) or serum (all others), pg/ml§      
 IFNγ456 ± 55.1691 ± 44.70.008762.3 ± 63.8509.2 ± 36.30.001
 IL-4168 ± 17.6228 ± 11.50.010179.1 ± 12.0259.8 ± 16.20.001
 IL-19.76 ± 0.659.20 ± 0.33NS11.1 ± 0.47.07 ± 0.420.001
 TNFα420 ± 241136 ± 22.80.012131.3 ± 22.6142.4 ± 41.7NS
 IL-677.6 ± 18.490.8 ± 9.74NS37.0 ± 4.0149.9 ± 17.90.001
 IL-1018.0 ± 2.2734.6 ± 3.050.01938.8 ± 4.3523.7 ± 2.760.005
 IL-12723 ± 185457 ± 66.5NS396.4 ± 65.7616.6 ± 113.6NS
Figure 2.

Differences in immunologic parameters between female and male arthritic (+) and nonarthritic (−) mice from the (BALB/c × DBA/2)F2 population immunized with cartilage proteoglycan (PG) to induce arthritis. Among the 16 immunologic markers or calculated parameters measured in the 559 immunized animals, only those which showed significant differences are presented. A, Serum levels of autoantibodies (aAb) against mouse (self) PG. B, Ratio of mouse IgG1 to IgG2a autoantibodies (aG1/G2a). Levels of C, interferon-γ (IFNγ) and D, interleukin-4 (IL-4) were measured in 4-day supernatants of PG-stimulated spleen cell cultures. Levels of E, IL-10, F, IL-6, G, IL-1, and H, tumor necrosis factor α (TNFα) were measured in serum. Values are the mean ± SEM. = P < 0.05; ∗∗ = P < 0.005, by Student's 2-sample t-test.

This was even more evident when arthritic females were compared with arthritic males or when nonarthritic females were compared with nonarthritic males (Figure 2). Serum IL-1 concentrations were significantly higher in female mice (either arthritic or nonarthritic) (Figure 2G), and IL-10 concentrations were significantly higher in nonarthritic female mice (Figure 2E). IFNγ production was more prominent in the nonarthritic female group than in either of the male groups (Figure 2C). IL-4 was significantly higher in both the arthritic and nonarthritic male groups than in the female groups (Figure 2D), whereas serum levels of TNFα were uniformly elevated in all arthritic mice compared with nonarthritic mice (Figure 2H).

Genome-wide linkage search for arthritis susceptibility genes.

In previous studies (34, 35), we used the arthritis score as a single trait, which was applied to both arthritic and nonarthritic mice. Using this definition of the arthritis trait, we identified 12 Pgia loci in a previous study (34). Although the experimental design was similar, the mapping was accomplished with 106 polymorphic markers and a different F2 population. This arthritis score, however, contained a mixture of several clinical traits. We therefore further divided the arthritis score into 3 independent scores: susceptibility to PGIA (binary), onset of the disease (onset), and severity of inflammation (severity). Although separation of clinical traits did not create biases among all traits, this step seemed to be a necessary procedure for correct calculations and linkage analysis of genes that might control the different features of arthritis. Indeed, differences between the 3 clinical traits (binary, onset, and severity) and their linkage to different Pgia loci clearly indicated the necessity of this approach.

The binary trait of disease susceptibility was mapped to chromosomes 7, 11, and 13 (Figure 3). The highest QTL was on chromosome 7 (LOD score 4.8), while the QTL on chromosome 13 showed an LOD score of 4.1 and the QTL on chromosome 11 had an LOD score of 3.1. All binary QTLs were high in females, but were lower or absent in males (Figure 3).

Figure 3.

Linkage analysis of clinical traits in a population of (BALB/c × DBA/2)F2 mice immunized with cartilage proteoglycan (PG) to induce arthritis. Quantitative trait loci (QTLs) were calculated for males, females, and both sexes together. The “arthritis score” used in previous studies (34, 35) was separated into subtraits: susceptibility to disease (binary trait), onset of disease, and severity of inflammation. The y-axes show logarithm of odds (LOD) scores, calculated using the free regression model of linkage for each chromosome. Whiskers along the x-axes show the positions of genomic markers according to the Kosambi linkage map for each chromosome. Interval mapping linkage analysis was performed using Map Manager QTX software. Table 2 shows information on peak marker positions (additional information is available from the authors upon request) and a summary of the QTLs for clinical and immunologic traits. New QTLs (Pgia26Pgia29) are indicated by asterisks. Loci with LOD scores exceeding the threshold of empirical significance established with the permutation test are boldfaced. Pgia28* on chromosome 11 is an immune-related QTL in F2 females without any evident correlation with clinical traits. hG1/G2a = ratio of human IgG1 to IgG2a autoantibodies.

Disease onset as a trait correlated significantly with arthritis, showing overlapping QTLs on chromosome 7 for disease onset (Pgia21 locus, LOD 2.5–3.0) and susceptibility (LOD 4.8). For all other chromosomes, onset was an independent trait from the binary or severity QTLs. The major QTL for onset was on chromosome 3 for females (LOD 4.9), and 2 smaller QTLs were found on chromosomes 10 and X (Figure 3 and Table 2).

Table 2. Summary of QTLs for PGIA in (BALB/c × DBA/2)F2 intercross mice*
ChromosomeQTLPosition, cMMarkerClinical traitsImmunologic traits
  • *

    Shown are the chromosomes that contain quantitative trait loci (QTLs) for clinical or immunologic traits of arthritis in major histocompatibility complex–matched (BALB/c × DBA/2)F2 hybrid mice, the position of the QTL (in cM), and the peak marker position. PGIA = proteoglycan-induced arthritis; LOD = logarithm of odds; TNFα = tumor necrosis factor α; IL-2 = interleukin-2.

  • Linkage for the binary trait, onset, and severity of the disease was calculated for males, females, and for the entire F2 population. The LOD score for each QTL is shown. Immunologic traits are those described in Table 1.

  • Loci with LOD scores exceeding the threshold of empirical significance established with the permutation test.

  • §

    New QTL identified in the present study. See references 34, 35, 37, and 38 for QTLs previously identified in PGIA.

1Pgia133–106D1Mit445Severity LOD 3.1, malesHetero IgG1:auto IgG1 LOD 3.7, males
3Pgia26§34–80D3Mit158Onset LOD 4.9, females
4Pgia130–20D4Mit97Severity LOD 3.1, malesTNFα LOD 5.1, males
4Pgia27§35–45D4Mit45Severity LOD 3.1, males
5Pgia180–10D5Mit1Severity LOD 3.3, malesHetero IgG1:IgG2a LOD 2.8, males
7Pgia2123–44D7Mit120Binary trait LOD 4.8, females; onset LOD 2.5, all F2
8Pgia41–22D8Mit224Severity LOD 3.9, all F2IL-2 LOD 3.5, females; hetero T cell proliferation LOD 3.5, females
9Pgia530–50D9Mit232Hetero IgG1:IgG2a LOD, 3.4 males
10Pgia630–40D10Mit40Onset LOD 3.5, females
11Pgia28§0–28D11Mit229Hetero IgG1:IgG2a LOD 2.8, males
11Pgia742–70D11Mit132Binary trait LOD 3.1, females
13Pgia150–47D13Mit179Binary trait LOD 4.1, femalesIL-1 LOD 3.1, females
14Pgia29§40–54D14Mit92Severity LOD 2.8, malesHetero T cell proliferation LOD 3.5, females
15Pgia80–20D15Mit252Severity LOD 3.5, females
15Pgia930–55D15Mit28Severity LOD 3.0, males
16Pgia1040–72D16Mit50IL-1 LOD 3.5, females
19Pgia1211–33D19Mit30Severity LOD 4.4, all F2
XPgia2567–72DXMit5Onset LOD 3.0, all F2Hetero IgG1:auto IgG1 LOD 3,4, all F2; IL-1 LOD 3.2, all F2; IL-6 LOD 3.5, all F2

The third clinical trait, severity, did not show any correlation with any of the other clinical traits. However, we found that a number of severity QTLs mapped to chromosomes 1, 4, 5, 8, 14, 15, and 19 (Figure 3 and Table 2).

When linkage analysis was performed for both sexes, all QTLs demonstrated either deviation of the peak position or different LOD scores for males and females. Interestingly, both binary and onset QTLs were higher in females than in males, and a few QTLs were fully sex-restricted, such as binary QTLs on chromosomes 11 and 13, and the onset QTL on chromosome 3 (Figure 3). In contrast, severity QTLs were more prominent in males than in females, showing several sex-restricted QTLs on chromosomes 1, 4, 14, and 15.

Comparison of clinical and immunologic traits and QTLs.

Table 2 summarizes the data from the QTL analyses of all clinical and immunologic traits that were scored in the entire F2 population and analyzed separately in males and females. Unexpectedly, only a few immunologic traits showed QTLs by linkage analysis. Seven of the immunologic traits showed overlap with clinical QTLs (Table 2), and most were also sex-restricted. Binary (susceptibility) QTL Pgia15 overlapped with the serum IL-1 QTL on chromosome 13. Disease onset QTL Pgia25 on the X chromosome was colocalized with the QTL for serum IL-1 and IL-6 and for PG-specific antibodies of IgG1 isotype (Table 2).

In general, most immunologic QTLs showed overlap with the severity QTLs, particularly at loci Pgia1, Pgia13, Pgia18, Pgia4, and Pgia29 on chromosomes 1, 4, 5, 8, and 14, respectively (Table 2). Some immunologic QTLs, however, did not match any clinical QTLs. For example, immunologic traits mapped to Pgia5 on chromosome 9, Pgia28 on chromosome 11, and Pgia10 on chromosome 16 did not show any relationship with clinical QTLs (Table 2).

Interaction between loci in different chromosomes.

We identified 4 loci (Pgia6, Pgia21, Pgia25, and Pgia26) that primarily contributed to disease onset (Figure 3 and Table 2), one of which (Pgia25) was on the X chromosome. Taking advantage of the large number of animals and the dense set of genomic markers used, we sought to determine possible interactions between onset QTLs (genes) on chromosomes 3, 7, 10, and X (Figure 4). We examined pairwise interactions (QTL to QTL), assuming that disease onset is affected not only by an allele represented by a single QTL peak marker, but also by a gene from another chromosome. Interaction was tested using the peak marker on the X chromosome (DXMit5 for QTL Pgia25) and peak markers D3Mit158 (Pgia26), D7Mit120 (Pgia21), and D10Mit40 (Pgia6). Allele–allele interactions were verified using chi-square statistics with 3 independent variables and a threshold P value of 0.05.

Figure 4.

Interaction of the onset quantitative trait locus (QTL) on the X chromosome (peak marker DXMit5) with onset QTLs on chromosomes 3, 7, and 10. The genotype of peak markers for these chromosomes was determined, and the onset score was calculated for each subset of mice carrying certain marker–marker combinations. A, Interaction of the X chromosome with chromosome 3 demonstrates that the source of arthritic allele is the BALB/c (B) parent and that the DBA/2 allele (D) is recessive in the context of the same (D) allele. In the heterozygous state (H), the alleles were codominant. B, Interaction of the X chromosome with chromosome 7 demonstrates that the arthritic allele is derived from the DBA/2 background. C, Interaction of the X chromosome with chromosome 10 demonstrates locus–locus interactions (i.e., the genotype at the X chromosome affects allele–allele interaction on chromosome 10). Values are the mean ± SEM. Asterisks indicate a significant difference between the 3 genotypes when the X chromosome was used as a reference genotype, by chi-square test. D, Further analysis of the allele–allele interaction shown in C demonstrates that the incidence of arthritis is under the influence of both chromosomes. Values are the number of arthritic mice/total number of mice and the incidence.

Figure 4A demonstrates the interactions between D3Mit158 and DXMit5 peak markers. The BALB/c allele of D3Mit158 (Bchr3) is dominant for disease onset, while the DBA/2 allele on chromosome 3 (Dchr3) is recessive, as suggested by a high onset score for the B allele and a low score for the D allele. In the heterozygous state, the DBA/2 and BALB/c allele interaction (Hchr3) is clearly codominant, resulting in intermediate onset scores. Dchr3–Hchr3–Bchr3 interaction (Pgia26) is not affected by Pgia 25 on the X chromosome; thus, allele Bchr3 is always codominant with allele Dchr3 (Figure 4A).

The same approach was applied to chromosomes 7 (Figure 4B) and 10 (Figure 4C). As in the case of chromosome 3, no interaction was found between chromosomes 7 and X. However, the gene locus marked with D10Mit40 on chromosome 10 was affected by the X chromosome (marker DXMit5), since mice heterozygous for the D10Mit40 locus and having DBA/2 alleles of DXMit5 (Hchr10–DchrX) showed significantly lower onset scores (Figure 4C, second column). To determine whether the susceptibility of arthritis was also affected by interacting loci on chromosomes X and 10, we compared the effects of loci represented by DXMit5 and D10Mit40 on the incidence of disease (Figure 4D). Indeed, incidence was affected in the same manner as onset, suggesting that both disease incidence and disease onset are under the control of the QTLs localized at chromosomes X (Pgia25) and 10 (Pgia6).

DISCUSSION

Immunization of BALB/c mice with human cartilage proteoglycan induces progressive autoimmune polyarthritis, which leads to complete deterioration of the articular cartilage and to joint deformities (28, 33), as in RA. There is strong genetic linkage between the MHC and autoimmune/arthritic processes (37); however, having the “right” combination of MHC alleles is not sufficient for the induction of an autoimmune disease (29, 35, 44, 53). In order to exclude the MHC effect on disease development and severity in a mixed (F2) genetic background, we intercrossed PGIA-susceptible BALB/c females with arthritis-resistant DBA/2 males, both strains carrying the same H-2d haplotype. The F1 progeny were further mated to generate (BALB/c × DBA/2)F2 hybrid mice, which were then immunized for PGIA. As expected, no disease-linked QTL was identified at MHC loci on chromosome 17. Since the original MHC (H-2d) function in this particular intercross was unchanged, the PG epitope repertoire was presented and recognized by autoreactive T cells in a similar manner in F2 hybrids and the PGIA-susceptible parental BALB/c strain. Therefore, this particular genetic combination allowed us to examine interactions between QTLs without the effect of the MHC. Among these loci, we identified 4 new QTLs (Pgia26Pgia29), all of which were clearly affected by sex (Figure 3 and Table 2).

Statistical estimates for LOD score thresholds that were modeled by Lander and Kruglyak in 1995 (50) were based on a theoretical murine population with a total genome size of 1,600 cM. However, since each F2 hybrid population differs by the distributions of markers and traits and biases between them, we established empirical LOD score thresholds using a permutation test (49, 51). Taking into consideration that permutations require some computational power, we considered for further calculations only traits that showed linkage above the theoretically “suggestive” LOD score of 2.8 (50). These were binary and onset traits in females and severity traits in males and females. To determine empirical threshold values for mapping these traits, we used 1,000 permutations, which were analyzed genome-wide for all 20 genotyped chromosomes in 1-cM intervals. Permutation analysis allowed us to establish suggestive, significant, and highly significant levels for QTL detection, corresponding to the 37th, 95th, and 99.9th percentiles, respectively.

As a result of this analysis, suggestive levels for all clinical traits were even lower than the theoretical estimates, with an empirical LOD score of 2.3 versus a theoretical LOD score of 2.8 (50). Significant empirical thresholds (LOD scores of 4.1 and 3.9 for binary and severity traits, respectively) were milder than the theoretical estimation (LOD score 4.3 [50]). A significant threshold for disease onset in females was established at 4.9, which is noticeably higher than the theoretical level. Therefore, as a result of permutation analysis, only 5 QTLs met the criteria for a significant empirical threshold with residual error probability of P < 0.05 genome-wide. These are binary QTLs Pgia15 and Pgia21, onset QTL Pgia26, and severity QTLs Pgia4 and Pgia12 (Table 2 and Figure 3). Thus, applying these more stringent criteria, we found 1–2 major loci for each independent clinical trait of the disease in this study. We did not find any colocalization or overlapping between these major loci/gene sets.

One of the most important findings of this study was that disease onset, as a clinical trait, was significantly different between F2 males and females (Figure 1). Linkage analysis explored numerous QTLs, which were influenced by sex and were linked to either a binary (susceptibility) or a quantitative (onset or severity) trait. Disease susceptibility and onset showed predominant linkage to the female sex, while the severity QTLs were prevalent in males. Only 1 QTL (Pgia21 on chromosome 7) was involved in both susceptibility and onset, whereas all other loci seemed to control only a single clinical trait in PGIA. Occasionally, QTLs were also linked to immunologic traits.

In addition to the well-documented effect of the MHC (Pgia17) (excluded in this study), only 3 QTLs appeared to control disease susceptibility. These results corroborated the findings of our previous studies on the MHC (35, 37), further supporting our conclusion that arthritis-susceptibility QTLs/genes (Pgia7, Pgia15, Pgia17, and Pgia21) do not control the severity of inflammation.

The MHC was “invisible” in this genetic cross because it was intentionally excluded when H-2d haplotype-matched parental strains were used. However, even in the absence of MHC loci, the relationships between sex, immunologic traits, and genetic background proved to be very complex. Nongenetic and environmental factors occasionally play as important roles in susceptibility and disease severity as the genetic components. Among the nongenetic factors, the individual's sex and age seem to have the strongest influence on the development of disease.

In humans, the prevalence of RA is significantly higher in women than in men, especially at younger ages. Before the age of 39 years, the incidence of RA in women is 19 times higher than that in age-matched men, although this difference is only 2–3-fold higher by the age of 60 years and older (1). This exaggerated ratio of women to men, ∼2.5:1, has been described in many studies of RA patients (1, 16, 17). In contrast, case–control studies have shown that erosive destruction occurs more frequently in men than in women (72% versus 31%; P < 0.05), although joint deformities are more pronounced in women (17). Considering susceptibility to RA and severity of RA as 2 distinct features of a complex disease, we can conclude from clinical studies that both the incidence (or prevalence) and progression (severity, flares, complications) of RA are features affected by the sex of the subject. According to a study involving more than 500 patients, however, age-related differences between female and male patients with RA were not found (16).

In animal models of RA, differences in many aspects of arthritis between males and females have been demonstrated. Arthritis scores in (DA × ACI)F2 hybrid rats with CIA were significantly affected by sex (20), in particular, by locus Cia5 on rat chromosome 10 (21). The Oia3 locus in rats with oil-induced arthritis, which corresponds to Cia5, was also identified as a sex-affected locus (22). Similarly, sex-dependent variations in QTL penetrance were demonstrated in F2 hybrid and congenic rats with CIA or with oil-induced or pristane-induced arthritis (18, 23). The MRL/lpr lupus-prone mouse strain (carrying a mutated Fas gene) developed arthritis spontaneously, and QTLs on chromosomes 2 and 15 were both shown to be sex-affected (54).

The importance of nongenetic factors in disease development should not be overlooked. Since genetic risk factors can explain approximately one-half of the population variability of RA, disease heritability in animal models could be even higher. It is interesting that in RA, twins showed a reduced value of heritability ranging from 0.56 to 0.65 for both sexes (maximum value is 1.0), albeit the value was estimated as 0.83 for the male proband subsample and essentially zero (1 × 10–5) for the opposite sex (55, 56). Consequently, the role of nongenetic factors (environment, sex hormones, behavior, and age-related changes among them) in arthritis have at least the same importance as genetic factors, and lifestyle, age, certain working conditions, exercise, sex, exposure to silica, obesity, and smoking are among the well-known risk factors for RA (57–62). The importance of nongenetic factors, such as aggressive behavior or estrogen turnover, in several mouse models of arthritis has been demonstrated as well (63, 64).

In the PGIA model, parental BALB/c female mice are more susceptible to arthritis and develop arthritis faster than do males (28, 29). In the genetic cross that was used in the present study, significant differences in arthritis scores (both severity and binary), as well as in many of the immunologic parameters, were found between females and males. Almost one-half of the analyzed traits demonstrated differences between sexes (Table 2). In this experimental condition, the only source of the Y chromosome was the DBA/2 strain. Thus, in the F2 population, technically, we compared the effects of the X chromosome from BALB/c mice with the effects of the Y chromosome from DBA/2 mice.

Existing linkage analysis software was usually designed to find individual genetic determinants, which control different traits, and then examine possible interactions between them. This approach may exclude chromosomal loci that control disease but only in cooperation with another locus. Locus–locus interactions in arthritis have been reported for QTLs on chromosomes 1 and 2 in mice (42) and on chromosomes 4 and 10 in rats (65). Addressing the question of how sex affects susceptibility to PGIA and disease severity, we tested the interaction of the QTL on the X chromosome (Pgia25) with onset QTLs on other chromosomes. Loci D3Mit158 and D7Mit120 (on chromosomes 3 and 7, respectively) were clearly independent on the X chromosome (Figures 4A and B), although these loci showed linkage to disease: F2 mice carried an arthritis-susceptibility allele (allele B) on chromosome 3, and an allele on chromosome 7 (allele D).

The relationship between alleles was peculiar when the QTLs on chromosome 10 and on the X chromosome were correlated (Figure 4C). The onset score was significantly lower when heterozygosity at D10Mit40 was paired with the DBA/2 allele DXMit5, indicating a nonallelic interaction between onset genes on the 2 chromosomes. This analysis was performed separately for males and females as well as for the combined population, and the pattern of interaction was basically the same. Therefore, we presented data for the total population because we were able to use a 2 times greater number of animals for determining significance levels and because DXMit5-heterozygous mice exist in the female subpopulation only.

Interaction between the X chromosome and chromosome 10 indicated that these 2 arthritis-related genes should operate in the same or related pathophysiologic processes. Considering genomic markers with LOD scores >2.0 as flanking markers for the QTL, we were able to locate the QTL on chromosome 10 within a 10-cM region (between 30 and 40 cM), and a QTL on the X chromosome within a 5-cM region (between 67 and 72 cM) (Table 2). According to the Celera Discovery System (52) and public databases (66, 67), each chromosomal region contains ∼150 genes, and the functions of at least half of these genes are not yet known. Screening through the list of 150 transcripts in each region, we found at least 10 known genes that are likely to be contributors to the pathogenesis of arthritis, since their functions are related to the regulation of cell growth, apoptosis, cell–cell adhesion, or the production of transcription factors that trigger inflammatory responses. Using congenic strain studies is an evident next step toward narrowing the chromosomal region and/or testing locus–locus interactions.

Table 2 summarizes data on the QTL analyses for all immunologic traits that were scored separately in the F2 population for males and females that developed PGIA. One could consider those QTLs as putative loci for PGIA, even if they did not show colocalization with arthritis QTLs in each case. Numerous immunologic QTLs overlap with clinical QTLs, thus providing information about possible mechanisms that underlie the function of the QTLs. The value of these data can be further increased by collecting information on positional gene candidates, which are now available in public databases (66, 67) and the Celera Discovery System (52). By projecting the 3 maps of positional gene candidates, clinical QTLs, and immunologic QTLs onto the genome, we could greatly accelerate the process of unraveling disease-associated genes.

Acknowledgements

We thank Dr. Kenneth Manly (Roswell Park Cancer Institute, Buffalo, NY) for discussion of linkage analysis issues and for help in clarifying critical features of the Map Manager QTX package. We thank Dr. Jeffrey M. Otto (Genaissance Pharmaceutical, New Haven, CT) for sharing his expertise on linkage analysis. We greatly appreciate the expert assistance of Sonja Velins (Rush University, Chicago, IL) for preparing the manuscript.

Ancillary