Disease-promoting and -protective genomic loci on mouse chromosomes 3 and 19 control the incidence and severity of autoimmune arthritis

Abstract

Proteoglycan (PG)-induced arthritis (PGIA) is a murine model of rheumatoid arthritis. Arthritis-prone BALB/c mice are 100% susceptible, whereas the major histocompatibility complex-matched DBA/2 strain is completely resistant to PGIA. To reduce the size of the disease-suppressive loci for sequencing and to find causative genes of arthritis, we created a set of BALB/c.DBA/ 2-congenic/subcongenic strains carrying DBA/2 genomic intervals overlapping the entire Pgia26 locus on chromosome 3 (chr3) and Pgia23/Pgia12 loci on chr19 in the arthritis-susceptible BALB/c background. Upon immunization of these subcongenic strains and their wild-type (BALB/c) littermates, we identified a major Pgia26a sublocus on chr3 that suppressed disease onset, incidence and severity via controlling the complex trait of T-cell responses. The region was reduced to 3 Mbp (11.8 Mbp with flanking regions) in size and contained gene(s) influencing the production of a number of proinflammatory cytokines. Additionally, two independent loci (Pgia26b and Pgia26c) suppressed the clinical scores of arthritis. The Pgia23 locus (~3 Mbp in size) on chr19 reduced arthritis susceptibility and onset, and the Pgia12 locus (6 Mbp) associated with low arthritis severity. Thus, we have reached the critical sizes of arthritis-associated genomic loci on mouse chr3 and chr19, which are ready for high-throughput sequencing of genomic DNA.

INTRODUCTION

Despite the progress in modern genomics and proteomics, which have enabled us to acquire knowledge of virtually all genes in a dozen species, finding a primary gene involved in a multifactorial complex disease, such as autoimmune arthritis, is still a daunting task and an enormous challenge. Rheumatoid arthritis (RA) is a chronic autoimmune disease affecting up to 1% of the human population. The major genetic risk factor for RA is the histocompatibility complex (MHC) on human chromosome 6, whereas other genetic alleles account for the remaining portion of the RA population whose genetic variances are not linked to MHC, and these alleles are distributed over the human genome.^{1– 4} Since the late 1990s, over 5500 RA patients and over 20 000 controls have been tested with HapMap SNPs (single-nucleotide polymorphism), and over 30 non-MHC RA risk alleles have been identified.^{4– 6} Different genomic loci and genes are linked to RA in different ethnic populations,^{4,5,7– 10} and some of these SNP-identified genomic alleles show very high correlation with the disease and seropositivity of RA.⁴ However, only a very few of these risk alleles represent genetic defects, but rather indicate a chromosomal region where disease-promoting genetic mutations or epigenetic alterations should be present.¹¹

The use of rodent models for RA research was initially intended to provide a less sophisticated, but faster and more easily manipulated system for identifying arthritis-susceptibility genes. The feasibility of the use of rodents was supported by the high homology between human and rodent genomes, and the very frequent and high similarities in the syntenic chromosome regions controlling arthritis in different rodent models and in humans.^{12– 16} These two fundamental findings supported the hypothesis that some of the major pathophysiologic pathways of arthritis in humans and rodents may be very similar, possibly identical.

Proteoglycan (PG) aggrecan-induced arthritis (PGIA) is a murine model of RA.^{16– 19} PGIA is a pertinent RA model on the basis of similarities in joint histopathology, clinical features, recessive inheritance, seropositivity (both rheumatoid factor and anti-citrullinated antibodies) and last but not least, the dominant role of the MHC locus.²⁰ Overall, 12 Pgia loci that syntenic to 18 of the most important (dominant) RA-associated human genomic regions identified so far.^4,16 Like RA, PGIA is under the control of multiple non-MHC chromosome loci, and the major quantitative trait loci (QTLs) are localized on murine chromosomes (chr) 2, 3, 7, 15 and 19.^{21– 27} The arthritis-susceptible (arthritis-prone) BALB/c strain carries the same H-2d MHC allele that is present in the resistant DBA/2 strain. The original Pgia26 locus contains the region where the Ptpn22 gene is located (http://www.ensembl.org; NCBI m37 mouse assembly), and this gene has been implicated not only in RA, but also in other human autoimmune diseases (see review papers for details^{4,11,28– 31}). Both Pgia23/mCia12 and Pgia12 loci on chr19 (in three different genetic crosses^{21– 23}) confer increased resistance to PGIA, and these loci overlap with the RA loci on human chromosome 9 where a number of RA-associated traits are located.^4,5,9,32,33

We have chosen the Pgia26 (from 93 to 130 Mbp) and Pgia23/ Pgia12 loci (overlapping the entire chr19) in congenic strains (Figure 1) for further genetic mapping of arthritis-susceptibility genes because these QTLs demonstrated major genetic effects in this model, and also in collagen-induced arthritis.^23,24,34 Moreover, as described above, these QTLs overlap (syntenic) with known RA loci in humans. We generated a set of BALB/c.DBA/2-congenic strains in which the PGIA-resistant loci from the DBA/2 strain were transferred to the susceptible BALB/c background.²⁶ We then selected mice with recombinations flanking the target intervals in order to reduce the size of the original QTLs, and to map the arthritis-susceptibility loci at higher resolution. We narrowed each mouse chromosome region to 3–6 Mbp in size (up to 19 Mbp with the flanking regions), which is small enough for high-throughput sequencing of genomic DNAs from arthritis-resistant and -susceptible strains.

Figure 1.

Genetic structure of the Pgia26 locus on chr3 (a) and the Pgia23/Pgia12 locus on chr19 (b) in subcongenic strains of BALB/ c.DBA/2 mice. Genomic loci of DBA/2 origin were transferred to the BALB/c genetic background, and speed congenic backcrossing was performed with genomic markers located within the originally identified large chr3 and chr19 loci, flanked with D3Mit311 and D3Mit323 markers for the Pgia26 locus, and D19Mit95 and D19Mit123 markers for the Pgia23/Pgia12 locus. Later, higher density marker sets were used to determine more detailed intervals between DBA/2- and BALB/c-homozygous regions shown with solid vertical lines (BALB/c) with flanking DBA/2 intervals (gray closed boxes along chromosomes). Sets of congenic/subcongenic strains used for genomic mapping include seven strains for both chr3 and chr19. The marker positions are shown in million nucleotide base-pairs (Mbp) projected along the chromosomes (left side). Order and position of the markers are in compliance with the mouse genome database (http://www.ensembl.org; NCBI m37 mouse assembly). Using this set of congenic strains, two arthritis-controlling loci (Pgia26a, Pgia26b) and one sublocus affecting an array of immune parameters (Pgia26c) were identified on chr3. The originally overlapping Pgia23 and Pgia12 loci (Pgia23/Pgia12 locus) on chr19 were separated and narrowed to an interval flanked by D19Mit128 and D19Mit16 genomic markers (Pgia23) and D19Mit30 and D19Mit13 markers (Pgia12). The codes of the congenic/subcongenic strains are indicated on the top of each line, and correspond to the labeling used in Tables 1 and 2. Different loci are indicated at the right side of the markers.

RESULTS

Mapping of Pgia subloci in congenic strains

Initially, in the F2 generation of the intercrossed arthritis-resistant DBA/2 and arthritis-susceptible BALB/c mouse strains, the major locus (Pgia26) controlling PGIA onset and severity was found on chr3,^27,35 and the locus controlling PGIA severity was mapped to chr19 (Pgia23/Pgia12).^26,27 Chromosomal positions of these loci were confirmed in congenic strains.²⁶ The initial size of the Pgia26 locus was 60 Mbp of the DBA/2-homozygous ‘core,’ and it could have been larger (up to 64 Mbp) if flanking BALB/c-homozygous areas were taken into consideration (Figure 1; Table 1). Similarly, the Pgia23/Pgia12 locus was 46 Mbp in size, 58 Mbp with the flanking regions. Assuming a random distribution of genomic markers within each locus, six subcongenic strains might provide a genomic resolution of 7.5–10 Mbp on both chromosomes. However, we did not select the genomic markers randomly in the breeding of these congenic mice, rather on the basis of the detailed structure of the loci obtained from marker regression analysis of the (BALB/c × DBA/2) F2 population (Figures 3a and e, dotted black curves).

Table 1.

Summary of genomic regions of chr3 tested for arthritis and corresponding data of incidence, severity and onset

Markers and their genomic positions in mega basepair (Mbp) are indicated. Shaded areas represent DBA/2 (D) chromosome intervals, and white areas with ‘B’ represent BALB/c alleles. Boxes indicate positions of disease-protective genomic intervals carrying DBA/2 alleles that suppress the incidence, severity and onset of arthritis (promoting in BALB/c). Data are presented as mean (for Incidence), or mean ± s.e.m. for arthritis and onset scores. Arthritis scores were recorded at the end of the observation period (days 59– 65). Difference of means between chr3 subcongenic and BALB/c mice was calculated using Student’s t-test assuming equal variances after Levene’s test for equality of variances. Trait Incidence was analyzed with Pearson χ² test, and trait Onset Score was analyzed with Mann– Whitney U-test. The major arthritis-protective Pgia26a and Pgia26 intervals are located between 108.8 and 118.2 Mbp position and flanking markers D3Abn158 and D3Mit142 weaker protective Pgia26c is located between 131.7 and 156.5 Mbp and flanked with D3Mit110 and D3Mit89 genomic markers. Levels of statistical significance for the comparison of congenic versus BALB/c mice are indicated with asterisks:

^*P < 0.05,

^**P < 0.001

^***P < 0.0005 and bold-faced.

Figure 3.

Linkage analysis for clinical and immune parameters in chr3- and chr19-subcongenic strains. The detailed structures of the Pgia26 locus on chr3 and the Pgia23 locus on chr19 were confirmed using three independent approaches. First, locus positions were established based upon the comparison of congenic strain genetic structures, and the corresponding clinical phenotypes for these strains were determined upon immunization (Tables 1 and 2). Two arthritis-susceptibility loci, Pgia26a and Pgia26b on chr3, and Pgia23 on chr19 were identified using the data in Table 3. Second, we analyzed (BALB/c × DBA/2) F2 population (n = 230) using a marker-by-marker χ² test (dotted black curves on the top panels ‘a’ and ‘e’). Third, a similar marker-by-marker approach was applied for the clinical and immunological traits of congenic mice (panels ‘b–d’ for chr3 and panels ‘f–h’ for chr19) with the assumption that all mice of these congenic strains and WT BALB/c mice belonged to one combined general population. Associations between traits and genotypes are presented as negative logarithms of P-values after Mann–Whitney U-test [–Lg (P-value) on y axis]. Only traits that showed an association greater than P < 0.01 [–Lg (P-value) >2] are presented.

Arthritis phenotypes and gender association in narrowed loci of subcongenic strains

To find prominent (major causative) genes controlling arthritis in these congenic strains, we immunized 25–63 mice of each subcongenic strain and collected data for disease characteristics, such as onset, incidence and severity of arthritis (Figure 2; Tables 1 and 2). Since Pgia26 was found initially to be a ‘female-specific’ locus, and the Pgia12 locus was ‘male-specific’ when linkage analysis of the F2 hybrid population was performed,^26,27 we immunized both females and males of the congenic/subcongenic strains along with wild-type (WT) BALB/c females (n = 87) and males (n = 54). When all of the data from the clinical phenotypes of all WT, congenic and subcongenic strains were analyzed together (389 total animals for chr3 and 357 for chr19), the gender effect of both loci was found to be negligible.

Figure 2.

Incidence, time of onset and severity of arthritis phenotypes in chr3- and chr19-congenic strains. Immunization revealed differences in arthritis susceptibility (upper two panels: a and c) and severity (lower two panels: b and d) of arthritis between subcongenic strains carrying different parts of the original large (Pgia26 and Pgia23/Pgia12) loci. One congenic and six subcongenic strains (with age- and gender-matched WT littermates) of both chromosomes were immunized for PGIA (Figure 1), but only results for strains with statistically significant genetic effects are shown. Although subcongenic c3A females showed positive gender-related differences, it did not reach a significant level when compared with WT females (data not shown). All other chr3-subcongenic males and females were highly comparable and, therefore, the data for all were combined. The chr3 loci influenced the onset, incidence (a) and severity (b) of arthritis. In addition to the original congenic line (c3-Pgia26, filled dark blue rectangles), the most substantial arthritis-suppressive effect was found in strains c3C (light blue triangles) and c3D (green-filled circles). The two loci (Table 2) on chr19 controlled both disease severity (c) and onset, but less extensively the susceptibility (shown as incidence c). Arthritis severity in c19C-subcongenic mice (green closed circles; (d)) was only 30% of that observed in WT BALB/c mice (red solid diamonds). Statistical differences between strains (color-coded) were calculated using Student’s t-test for the last day of observation (days 59–65) as shown in Table 1 and indicated on each panel.

Table 2.

Summary of genomic regions of chr19 tested for arthritis and corresponding data of incidence, severity and onset

Markers and their genomic positions in mega basepair (Mbp) are shown. Shaded areas represent DBA/2 (D) chromosome intervals, and white boxes with ‘B’ represent BALB/c alleles. Thick-framed areas correspond to arthritis-protective intervals, which are either reduces onset and incidence of arthritis (peak marker D19Mit128; Pgia23) or suppresses arthritis severity (peak marker D19Mit13; Pgia12). Data are presented as mean (for Incidence), or mean ± s.e.m. for arthritis and onset scores. Arthritis scores were recorded at the end of the observation period (days 59– 65) as described in Table 1 legend and Materials and methods. Difference of means between chr19 subcongenic and BALB/c mice was calculated using Student’s t-test assuming equal variances after Levene’s test for equality of variances. Trait Incidence was analyzed with Pearson χ² test, and trait Onset Score was analyzed with Mann– Whitney U-test. The major arthritis-protective Pgia23 interval is located between 14.00 and 20.42 Mbp position up to the flanking markers D19Mit95 and D19Mit14, and Pgia12 protective is located between 26.88 and 33.01 Mbp flanked with D19Mit80 and D19Mit88 genomic markers. Levels of statistical significance for the comparison of congenic versus BALB/c mice are indicated with asterisks:

^*P < 0.05,

^**P < 0.001

^***P < 0.0005 and bold-faced.

Arthritis phenotypes in congenic/subcongenic strains of chr3

Arthritis severity was significantly suppressed by 50% (P < 0.05) in the original c3-Pgia26-congenic strain (Pgia26 locus) as compared with the WT BALB/c littermates (Table 1; Figure 2b). By the last week of observation (days 59–65), only 60% of congenic c3-Pgia26 mice developed arthritis, whereas the onset of these arthritic mice (first visually detectable sign of inflammation) was similar to their WT BALB/c littermates (males and females are together) (Figure 2a). Two subcongenic strains (c3C and c3D) demonstrated a dramatic suppression of inflammation (confirmed by histology, data not shown) with significantly delayed onset and a very low incidence of arthritis (P < 0.0005) (Figures 2a and b; Table 1). Subcongenic mice (c3D) carrying the DBA2 allele of chr3 from the 99.9 to 131.7 Mbp region were almost completely resistant to PGIA: 0.97–1.07 arthritis score with 39% or less incidence (Figures 2a and b; Table 1). The protective effect of DBA/2 alleles in different subcongenic strains correlated directly with serum levels of interleukin (IL)-1β and tumor necrosis factor (TNF)-α , both of which were not measurable in the subcongenic c3C and c3D animals. All other cytokines and autoantibodies were highly comparable in each subcongenic strain measured at the end of the experiment (data not shown).

Analysis of the clinical phenotypes suggested that the critical region controlling the resistant phenotype, and that provided almost full protection against PGIA in c3C and c3D subcongenic strains, was located between markers D3Mit158 and D3Mit142 (between D3Mit103 and D3Mit316 (107.2–120.4 Mbp) with the flanking regions). The genetic effects of smaller or different DBA/2 intervals in the c3B, c3E and c3F subcongenic strains were clearly detectable, but far less extensive than in the c3C and c3D mice (Table 1). Subcongenic strain c3A was indistinguishable from WT BALB/c mice, thus the disease-protective (suppressive) region could be excluded from telomeric of marker D3Mit215. Subcongenic c3E and c3F strains were also close to WT littermates, and all three clinical phenotypes (incidence, onset and severity) of arthritis were only 15–20% lower in these subcongenic mice than in WT BALB/c mice (Table 1). Analyzing all of the clinical phenotypes in all congenic/subcongenic strains, we concluded that the major arthritis-protective region from DBA2 was located telomeric of the marker D3Mit158 position but ‘centromeric’ of marker D3Mit215 (between the 108.8 and 120.4 Mbp including the flanking regions), and we designated this region the Pgia26a locus (Figure 1a). Either a similar or a weaker protective region ‘centromeric’ of 112.8 Mbp (D3Ds112) marker position (between 107.2 and 115.8 Mbp including the flanking regions) was supposed to be present, and this region was designated the Pgia26b locus (Figure 1a). Taken together, the two major non-MHC disease-associated loci (protective in DBA/2 and disease promoting in BALB/c strains) were located between the D3Mit103 and D3Mit316 marker positions within a region that was < 14 Mbp. The relatively mild effect of the DBA/2 allele in the subcongenic c3E and c3F mice was most likely due to the combination of (D3Mit101–D3Mit199) and disease-protective DBA/2 allele located between the 131.6 and 165.5 Mbp position (D3Mit110–D3Mit89, including the flanking regions) on chr3 (Pgia26c) (Figure 1a).

Arthritis phenotypes in congenic/subcongenic strains of chr19

In addition to the original c19-Pgia23/Pgia12-congenic line,²⁶ four additional subcongenic strains demonstrated significant reductions in inflammation in peripheral joints when compared with the BALB/c controls (Figures 2c and d; Table 2). One of these four subcongenic strains (c19C) exhibited a very similar phenotypic profile as the parental congenic strain (c19-Pgia23/Pgia12) (Table 2). The genetic effect of Pgia12 locus on chr19 was exclusively related to arthritis severity in this congenic (c19-Pgia23/Pgia12) and the four subcongenic (c19C–F) lines. In the parent congenic strain and subcongenic line c19C, all clinical phenotypes of arthritis were significantly affected (P < 0.0005) (Figures 2c and d; Table 2). In contrast to the four subcongenic strains (c19C–F), severity was only minimally affected in c19A and c19B subcongenic lines, but the reduced incidence and late onset of arthritis was significantly lower in these mice than in the WT BALB/c littermates (Table 2). Because at least one or all three clinical phenotypes in all congenic and subcongenic strains were significantly affected on chr19, there were remarkable differences among subcongenic strains. The onset and incidence of arthritis was controlled by the region where the peak marker was D19Mit128 (between 14.0 and 20.4 Mbp region), whereas severity was suppressed by gene(s) located at peak marker D19Mit13, which may include a maximum length of a DBA/2 allele between 26.88 and 37.33 Mbp position (Table 2). Reanalyzing all previous genome-wide association studies performed earlier on the 1292 arthritic mice out of the 3234 immunized F2 hybrid mice,¹⁶ and reanalyzing the 29 Pgia QTLs, we maintained the original nomenclature of the reduced size (3.3 Mbp) of Pgia23 locus for arthritis onset and susceptibility, and the Pgia12 locus (< 10.5 Mbp) dominant for arthritis severity, both on chr19.

Fine mapping of PGIA loci using marker regression analysis for clinical traits

Strain distribution pattern analysis and strain-by-strain comparison are straightforward methods for mapping subloci within a known chromosomal interval.^36,37 Therefore, we were compelled to further evaluate and corroborate the results using two additional approaches. First, we reanalyzed (BALB/c ×DBA/2) F2 hybrid mice from an earlier experiment²¹ using single marker regression analysis, and then we compared the F2 population-based results with the most recent data (Figures 3a and e). Second, we performed single marker regression analysis for the combined population of congenic mice (total n = 389 for chr3, and total n = 357 for chr19). In the latter case, mice were not grouped by congenic strains but were assumed to be part of a non-segregating general population.^36,37

Marker-by-marker regression analysis for clinical traits, in the context of either BALB/c or the DBA/2 genotype, was performed (Figures 3a and e). Single marker regression analysis also confirmed the positions of the three major loci of Pgia26a–c on chr3, and Pgia23 and Pgia12 on chr19, which were identified previously by strain distribution pattern analysis. No other locus was found on chr3 or chr19. The present results with WT and congenic/subcongenic mouse strains are the summary of seven independent immunization experiments for both chr3 (n = 389) and chr19 (n = 357), and were also based on the re-evaluation of n = 230 of F2 hybrids.

Correlations among cytokine responses, serum autoantibody levels and genomic loci

Biomarkers provide an important basis for the understanding of disease mechanisms and, if a biomarker is closely associated with an arthritis phenotype, it allows us to use the biomarker to more precisely map multiple loci within a given chromosome interval. Employing a marker regression approach similar to that used for mapping clinical scores in these congenic strains, we examined the linkage of the chr3 and chr19 loci using the various cytokine parameters (Figures 3b–d and f–h). The cutoff level for statistical significance was set at P < 0.01 using the Mann–Whitney U-test (–lg (P-value) >2, as indicated in Figure 3).

From the measured serum parameters, high levels of interferon (IFN)-γ were associated with both the Pgia26a and Pgia26b subloci (Figures 3a and b), although the correlation with arthritis severity was negative: cytokine concentration was 68% higher in DBA/2-homozygous Pgia26a or Pgia26b allele carrying mice (showing reduced arthritis severity compared with BALB/c-homozygous mice (P < 0.0001) (Table 3). Conversely, we found a positive correlation between arthritis severity and serum levels of IgG1 antibodies and IL-1β and IL-17 cytokines, as the serum levels of these biomarkers were lower in mice carrying arthritis-suppressive DBA/2 alleles (Table 3). Similarly, a positive association between greater arthritis severity and the presence of BALB/c alleles was found in chr19-congenic strains for serum IL-1β and IgG1 auto-antibodies. DBA/2-homozygosity in congenic mice was associated with low serum IL-1β (32% reduction) and IgG1 autoantibody levels (24% reduction) (Table 3).

Table 3.

Immune parameters linked to major chr3 and chr19 loci

	Stimuli	Chr3:Pgia26			Chr19:Pgia23 and Pgia12
		BALB/c	DBA/2	Effect	BALB/c	DBA/2	Effect
Serum
IL-1β		111 ± 17	85 ± 21	−23%*	109 ± 14	75 ± 11	−32%*
IL-17		46 ± 13	15 ± 5.3	−67%*	38 ± 8	34 ± 7	−10% n.s.
IFN-γ		143 ± 10	240 ± 21	+68%****	85 ± 11	92 ± 16	+8.2% n.s.
IgG1 (mPG)		127 ± 7.4	102 ± 13	−20%**	109 ± 6.3	83 ± 4	−24%**
In vitro responses
Proliferation	PG	4.1 ± 0.4	1.9 ± 0.2	−54%****	4.1 ± 0.3	3.6 ± 0.3	−12% n.s.
IL-2	PG	3.5 ± 0.1	2.6 ± 0.2	−25%****	3.1 ± 0.2	2.4 ± 0.1	−22%**
IL-4	Basal	100 ± 13	183 ± 33	+82%*	352 ± 59	271 ± 50	−23%**
	PG	1726 ± 131	1318 ± 155	−24% n.s.	3401 ± 217	2022 ± 167	−41%****
IL-6	Basal	204 ± 23	520 ± 69	+155%****	444 ± 64	493 ± 87	+11% n.s.
	PG	1428 ± 146	1729 ± 170	+21%*	2089 ± 189	1140 ± 130	−45%****
IL-17	Basal	523 ± 60	1345 ± 384	+157%*	458 ± 56	767 ± 110	+67% n.s.
	PG	1751 ± 159	3052 ± 524	+74%*	1969 ± 212	1923 ± 239	−2.3% n.s.
TNF-α	Basal	182 ± 16	488 ± 118	+168%*	293 ± 29	264 ± 30	−10% n.s.
	PG	375 ± 22	716 ± 207	+91% n.s.	621 ± 125.0	635 ± 131	+2.2% n.s.
IFN-γ	Basal	3357 ± 816	4569 ± 954	+36%**	3268 ± 1075	2213 ± 450	−32% n.s.
	PG	10 980 ± 1769	7513 ± 1154	−32% n.s.	6393 ± 823	3777 ± 633	−41%**

Abbreviations: IFN, interferon; IL, interleukin; TNF, tumor necrosis factor. Immune parameters, which showed statistical difference between congenic mice versus BALB/c wild-type mice, are presented as mean ± s.e.m. Statistical significance according to Mann– Whitney U-test for the differences of immune responses between mice carrying BALB/c-homozygous versus DBA/2-homozygous alleles for each locus is presented. Pgia26b peak marker is D3Mit199; Pgia12a peak marker is D19Mit16 (Table 1). Details for statistical analysis are provided in ‘Materials and methods,’ section ‘Mapping of genetic loci controlling clinical and immunologic features of arthritis.’ Threshold for statistical significance was set at *P < 0.05; ‘n.s.,’ not significant. Human aggrecan proteoglycan (PG) was used both for immunization and for in vitro lymphocyte stimulation (stimuli), whereas ‘basal’ indicates no stimulation (i.e., the spontaneous in vitro cytokine release). T-cell proliferation and IL-2 production (measured by CTLL-2 assay) are expressed as stimulation indices. Concentration of cytokines in mouse serum and in conditioned media of lymphocytes expressed in pg ml ⁻¹. From the complete list of measured parameters, only immune biomarkers, which showed linkage to loci, are presented. ‘Effect,’ is a percent change between DBA/2-homozygous allele and wild-type BALB/c mice. Statistically strongest effects are bold-faced (**P < 0.01, ****P < 0.0001). Values in italics are not significant (n.s.), and statistically significant values are bold-faced.

Antigen (PG)-induced in vitro-measured lymphocyte responses were also useful immune parameters in this set of subcongenic strains. Due to the genetic effect of the major Pgia26a sublocus on chr3, lymphocyte proliferation in subcongenic mice was 54% less than that in BALB/c-homozygous mice. This lower T lymphocyte proliferation was associated with less IL-2 production, but higher levels of IL-6 and IL-17 (Table 3). Thus, some of the differences in IL-4, IL-6, IL-17, TNF-α and IFN-γ cytokine levels might be attributed to an antigen-independent basal distortion in the distribution or function of lymphocytes, possibly also controlled by gene(s) in the Pgia26a sublocus.

The Pgia23 and Pgiai12 loci on chr19 were analyzed together, and no association with PG-induced lymphocyte proliferation was found. However, when these two loci were analyzed separately, Pgiai23 (Figures 3e–h), which is likely a small region between 14.0 and 17.3 Mbp region (Table 2), controlled antigen-dependent production of IL-2, IL-4, IL-6 (Figure 3g) and perhaps IFN-γ (Table 3). In contrast, the presence of arthritis-suppressive DBA/2 alleles in the Pgia12 locus (26.88–37.33/39.96 Mbp) was associated with a decreased production of IL-2, IL-4, IL-6, IFN-γ and slightly increased antigen-specific T-cell proliferation (Figure 3g) associated with significantly higher levels of ‘baseline’ (spontaneous) secretions of IFN-γ and TNF-α (Figure 3h). This telomeric 27–60 Mbp interval of chr19 (Pgia12) demonstrated more significant control over lymphocyte proliferation than Pgia23 (Figure 3g).

We have found multiple, and often very strong, linkages of measured immune parameters with chromosome intervals that did not fall within arthritis-controlling loci. For example, a 135–160 Mbp interval on chr3 was linked to serum TNF-α (Figure 3b), and the 27–51 Mbp telomeric part of chr19 were linked to serum IgG1 autoantibody (Figure 3f). Sublocus (considered to be Pgia26c, corresponding subcongenic c3F region in Table 1 between markers D3Mit127 and D3Mit323 (Figure 1)), was associated with high serum levels of TNF-α (Figures 3b and d) and antigen-induced IL-2 production (Figure 3c). Sublocus Pgia12 was linked to PG-specific lymphocyte proliferation and the production of IL-2 (Figure 3g), and an antigen-independent in vitro secretion of TNF-α and IFN-γ (Figure 3h).

DISCUSSION

Genome-wide association and conventional linkage analysis studies provide a complex picture of the genome with all potential disease-associated loci, if sufficient numbers of disease-affected individuals and appropriate control groups are compared. Congenic/subcongenic approaches employ purposefully selected rodent populations, focusing on a given chromosomal region identified by linkage analysis. In this study, we used MHC-matched (H-2d) congenic strains in which the Pgia26 locus on chr3 and Pgia23/Pgia12 loci on chr19 in an arthritis-susceptible BALB/c mouse strain were replaced with the corresponding region from arthritis-resistant DBA/2 mice.²⁶ These two congenic strains were selected because both loci exhibited significant suppression of arthritis phenotypes²⁶ and they were syntenic with human chromosomal regions associated with RA.^{4,5,7,11,28– 33,38}

The clinical phenotype and biomarker analyses indicated that the original Pgia26 locus on chr3 incorporated more than one sublocus and more than one disease-affecting genes. Pgia26a together with the Pgia26b, both of which are disease-protective/ suppressive loci from the DBA/2 strain, resulted in the most effective disease suppression in subcongenic lines c3C and c3D upon PGIA. These subcongenic lines appeared to be almost completely resistant to arthritis in all of the clinical features measured, including onset, incidence and severity. Both subcongenic lines, especially c3D, demonstrated a dramatic suppression of both incidence and severity of arthritis (Figures 2a and b; Table 1). Consistent with a multiple gene hypotheses, potentially, an arthritis-protective/preventing locus telomeric of Pgia26a/ Pgia26b in the c3A strain, was removed from the original c3-Pgia26-congenic strain (Figure 1a), and this was demonstrated to have the greatest effect on disease promotion. In a reciprocal manner, this also suggests that the major disease-suppressive gene/locus is located ‘centromeric’ of marker D3Mit215 (115.8 Mbp). However, a number of chr3-subcongenic lines (c3E, c3F and probably the c3B) carry the disease-promoting Pgia26a and Pgia26b loci of BALB/c origin. In contrast, genes from the DBA/2 strain, in the same regions, especially in the original c3-Pgia26-congenic and in the c3C- and c3D-subcongenic lines exert disease-protective function, that is, these congenic/subcongenic lines are far less susceptible to PGIA than the c3A or other subcongenic lines (c3E, c3F and c3B). Therefore, we expect a third arthritis-protective locus (Pgia26c) between marker positions D3Mit127 and D3Mit323 (142.6–152.4 Mbp), which is present only in subcongenic lines c3E and c3F and the original c3-Pgia26-congenic strain. Although this locus is relatively dense in ‘candidate’ genes (for example, Col24a1, Hs2st1, Ifi44, Nexn, Ptgrf and so on), the Bcl10 (B-cell leukemia/lymphoma10, at 145.59 Mbp) appears to be involved in inflammation, most likely via B-cell activation and/or nuclear factor κB activation in various innate immune cells and B cells released from the marginal zones of the secondary lymphoid organs.³⁹

Interestingly, c3C or c3D subcongenic strains exhibited a significantly more extensive disease suppression than the original c3-Pgia26-congenic line (P < 0.001 between c3C/c3D and c3-Pgia26), whereas the original c3-Pgia26 carried the Pgia26c locus as well. This observation suggests that epistatic interactions^23,24,34,35 may exist among disease-suppressive genes in an arthritis-prone BALB/c background. Nonetheless, this question of complex epistatic gene interactions will be studied (after sequencing the suppressive regions of interest) in vivo in combined IVSC (interval-specific congenic) strains (for example, the c3C and c3F DBA2 loci together in BALB/c background).

There are a total of 60 protein-encoding genes (plus six members of the Amylase family) and 13 spliceosomal or small nuclear RNA (snRNA) in the Pgia26a and Pgia26b loci (including overlapping flanking regions of the two genomic regions). Among these genes, there are Vav3 (guanine nucleotide exchange factor, oncogene, 109.14 Mbp, position of marker D3Mit158), which is responsible for normal osteoclast function;⁴⁰ Extl2 (exostoses-like 2, 115.71 Mbp), which is responsible for multiple exostoses;⁴¹ and Slc35a3 (solute carrier family 35, member 3, 116.37 Mbp), which is a gene, that if mutated, is responsible for multiple vertebra malformation and fusion.⁴² All three of these genes are involved in either normal or pathological skeletal function in RA and ankylosing spondylitis, and in the corresponding animal models of arthritis and spondylitis.^16,17,43 In addition, this area (Pgia26a–b) contains Vcam1 (vascular cell adhesion molecule 1; 115.81 Mbp), which together with its ligands may play role in the pathogenesis of RA.^{44– 46} This region carries Vav3 (see above) which is involved in T cell receptor (TCR) signaling,⁴⁷ and Prmt6 (protein arginine methyltransferase 6; 110.05 Mbp), which controls global levels of histone H3R2me2a methylation.⁴⁸

Although the arthritis-affecting direct function of Ptpn22 could not be confirmed in this study, the position of the Ptpn22 gene at 103.7 Mbp is matched with the locus peak at 101 Mbp (Figure 3a, marker regression results in F2 hybrids) within the original Pgia26 locus on mouse chr3.^27,35 However, when this locus was transferred to a pure BALB/c background, the linkage vanished (Figure 3a), and no clinical phenotypes of PGIA were consistently affected in any of the subcongenic lines (Table 1). The ‘Ptpn22’ locus (< 10 Mbp in size) is a very complex genomic area carrying 128 protein-encoding genes, 21 spliceosomes and snRNAs and 43 pseudogenes or genes having no protein-encoding transcripts. PTPN22 (on human chr1) has been shown to be a major risk allele in type 1 diabetes, Graves’ disease, systemic lupus erythematosus, juvenile idiopathic arthritis, autoimmune Addison’s disease, Hashimoto’s thyroiditis and RA.^{4,11,28– 31} Therefore, this Ptpn22 region, which is one of the major risk alleles in RA, cannot be ignored in genomic DNA sequencing and in vivo studies with mouse chr3 (syntenic with multiple disease-associated alleles in human chr1). In addition to PTPN22/Ptpn22, there are a number of genes in this region encoding proteins involved in or that play critical roles in immune function, such as Vtcn1 (V-set domain containing T-cell activation inhibitor 1, at 100.63 Mbp), CD101 (at 100.74 Mbp), CD2 (at 101.08 Mbp), Lrig2 (a leucine-rich repeats and immunoglobulin-like domains 2, at 104.26 Mbp), Igsf3 (immuno-globulin superfamily, member 3, at 101.18 Mbp) and Tspan2 (tetraspanin 2, expressed only in T cells, at 102.54 Mbp).

The genetic effects in the chr19-subcongenic strains were consistent with the single-gene hypothesis in both regions, as all corresponding congenic strains demonstrated either reduced arthritis severity or incidence, or both (Table 2). When the three clinical phenotypes (onset, incidence and severity) were analyzed side-by-side, it was clear that chr19 carried two disease-protective alleles, both from DBA/2 strain, which were narrowed but clearly located in the positions of the original Pgia23 and Pgia12, whereas these two QTLs originally were identified in different genetic crosses.^16,21,23,24

The Pgia23 locus was reduced to as small as 3–4 Mbp (between 14.00 and 17.33 Mbp), and the Pgia12 locus to <10.5 Mbp in size (26.88–32.71 Mbp). Clearly, while the Pgia23 locus carries genes responsible for arthritis incidence and early onset, the Pgia12 locus controls disease severity. If both were present (for example in the original congenic or c19C subcongenic lines), a dramatic decrease/ suppression was observed in all clinical phenotypes. The reduced size of Pgia23 (~3 Mbp) contained as few as 10 protein-coding genes (and two snRNAs), and probably the only candidate gene here is Tle4 (transducin-like enhancer of split 4 at position 14.52–14.67 Mbp), a gene responsible for B-cell function/maturation, and one that can be selectively detected in the nuclei of pro-, pre- and matured B cells.⁴⁹ The Pgia12 locus includes 46 protein and 5 spliceosomes/snRNA-encoding genes. Thus, the Pgia12 locus is a complex area with a number of candidate genes. Within this 10–11 Mbp region, there were (i) regulatory factor X 3 (Rfx3), which influences human leukocyte antigen (HLA) class II expression and affect DNA methylation;⁵⁰ (ii) Janus kinase 2 (Jak2); (iii) CD274 and (iv) its ligand Pdcd1lg2 (programmed cell death 1 ligand 2), which are costimulatory molecules involved in T and B cell cooperation. Moreover, this locus contains IL-33, a cytokine known to be involved in acute inflammation in RA and in a number of arthritis models.^{51 – 54}

In summary, using a set of 14 congenic/subcongenic strains, we achieved significant progress in mapping murine arthritis-susceptibility loci and reducing locus sizes. We also showed that linkage analysis of large populations (>200 mice) and congenic strains (>600 mice) are both very useful approaches that enabled us to unravel the complex genetic structure of chromosomal loci, thereby facilitating understanding of the mechanism of disease regulation, even when specific genes within these loci are yet unknown. Moreover, arthritis-associated genomic loci on mouse chr3 and chr19 reached critical sizes, which are ready for high-throughput sequencing using genomic DNA samples from the two parent strains and from individuals of disease-protected and disease-promoted subcongenic lines. Simultaneously with the ongoing sequencing studies, we have selected 14 IVSC strains with smaller overlapping recombinations within the Pgia26a–c loci for in vivo functional (arthritis-susceptibility and severity) studies to support genomic alterations within this small chromosomal region.

MATERIALS AND METHODS

Animal breeding

All animal experiments were approved by the Institutional Animal Care and Use Committee (Rush University Medical Center, Chicago, IL, USA). Animals were maintained in a pathogen-free environment. A marker-assisted selection protocol for the speed of congenic breeding was used to generate QTL-specific congenic strains. As described earlier,²⁶ congenic BALB/c.DBA/2-Pgia26 (chr3) and BALB/c.DBA/2-Pgia23/Pgia12 (chr19) mice passed six marker-assisted backcrosses. In brief, PGIA-susceptible BALB/c female and PGIA-resistant DBA/2 male mice (National Cancer Institute, Kingston Colony, NY, USA) were mated and the F1 males were backcrossed to parental BALB/c strain to obtain the N2 backcross generation (n = 500). N2 males were genotyped with 230 genomic markers selected for detectable length polymorphism between the parental strains using information from the Mouse Genome Informatics database (http://www.informatics.jax.org). N2 males bearing DBA/2-type heterozygous QTLs on chr3 or chr19, and the highest number of BALB/c alleles on the rest of the genome were selected for the next backcross. Approximately 40 offspring males after each backcross were genotyped with seven QTL-specific markers, and additional markers were used for the genomic regions found to be heterozygous in the previous backcross. At backcross level N6, when the entire genome was BALB/c-homozygous except for the selected QTLs, N6 males and females of the same genotype were intercrossed generating N6F1 population of BALB/c.DBA/2-QTL (c3-Pgia26 and c19Pgia23/12) founders. We intercrossed these mice to generate the homozygous N6F2 population, which was immunized with PG to induce arthritis. Arthritis suppression was demonstrated in both congenic strains when compared with the WT BALB/c strain.²⁶ In the second cycle of the study, homozygous c3-Pgia26 and c19-Pgia23/12 congenic males²⁶ (Tables 1 and 2) were intercrossed with WT BALB/c females, and males heterozygous for the Pgia26 or Pgia23/12 regions were repeatedly backcrossed with WT females for selection of appropriate recombination. New recombinations were selected with allele-specific markers (Tables 1 and 2). Heterozygous males and females with the same recombinations (usually brothers and sisters) were intercrossed to select founders to establish subcongenic lines homozygous for the selected sublocus (Figure 1).

Antigen, immunization and assessment of arthritis

Cartilage was obtained from osteoarthritic patients undergoing knee joint replacement surgery. The use of human cartilage for PG isolation was approved by the Institutional Review Board of the Rush University Medical Center, Chicago, IL, USA. PG isolation was conducted as described previously.^55,56 Mice were immunized with human PG at 12– 16 weeks of age using a standard immunization protocol.^55,56 Mice were immunized with 100 μg of PG protein and adjuvant, as described,^55,56 on days 0, 21 and 42. Arthritis severity was determined using a visual scoring system based on the extent of swelling and redness of the fore and hind paws.^17,56 Animals were examined three times a week alternatively by two or three investigators in a completely ‘blind’ manner. Inflammation was scored from 0 to 4 for each paw, resulting in a cumulative arthritis (severity) score ranging from 0 to 16 for each animal.^17,21,23,56 Small swelling of the paw or small joints (for example, inter-phalangeal or metacarpo- or metatarsophalangeal joints) is recorded as score 1, evident swelling and redness over the entire paw is scored 2, extensive swelling (~40– 50% increase of the original ankle, wrist, paw diameters) is scored 3, whereas the maximum redness and swelling (twice bigger paw in diameters) of a paw, or ankylotized joint(s), is graded as score 4.^17,55,56 Typically, once a joint inflamed, the score is increasing, and joint deformities are evident in the chronic phase. In each case, when the clinical (visual) score of a given paw is maximum 1, histopathology is used to confirm the level of joint inflammation.^21,23,57 The incidence of disease was expressed as the percentage arthritic mice out of the total number of mice. Additionally, the earliest day of disease onset was used to characterize how quickly a mouse developed arthritis. Negative mice were assigned an onset score ‘0’ and mice that developed arthritis most quickly in WT BALB/c controls were assigned an onset score of ‘5.’

Measurement of lymphocyte responses, cytokine and immunoglobulin production

Serum levels of autoantibodies to mouse PG of IgG1 and IgG2a isotypes were measured using enzyme-linked immunosorbent assay (ELISA) as described previously.^16,23,56 Serum cytokines IFN-γ , IL-1β, IL-4, IL-6, IL-17 and TNF-α were determined using capture ELISA kits purchased from BD Bioscience (San Diego, CA, USA) or R&D Systems (Minneapolis, MN, USA) as described.^16,23,56 Antigen (PG)-specific lymphocyte responses were measured in spleen cell cultures in the presence or absence of 50 μg ml ⁻¹ PG antigen. PG-specific T-cell proliferation was assessed on day 5 by the incorporation of [³H]-thymidine and was expressed as the stimulation index, which is a ratio of incorporated [³H]-thymidine in PG-stimulated compared with non-stimulated cultures.⁵⁶ IL-2 production was measured in the supernatants of the same spleen cell cultures on day 3 by CTLL-2 assay, as previously described.^19,23,56 The in vitro production of IFN-γ , IL-2, IL-4, IL-6, IL-17 and TNF-α in PG-stimulated and non-stimulated spleen cell cultures was measured on day 4 using capture ELISA.

Mapping of genetic loci controlling clinical and immunologic phenotypes of arthritis

We employed three independent approaches to localize smaller chromosome intervals within the originally identified large Pgia26 locus on chr3 and Pgia23/Pgia12 loci on chr19. First, we examined the strain distribution pattern and performed an association with clinical traits to narrow the chromosomal regions.^36,37 This approach is based on a pairwise comparison of the onset score, incidence and severity of arthritis for each subcongenic and original congenic strain, relative to WT BALB/c mice, to find an association between the strains’ phenotypes and genotypes (Tables 1 and 2).

Second, we analyzed the data from the (BALB/c × DBA/2) F2 combined population of males and females (n = 230) (shown in Figures 3a and e, dotted black curves). The same F2 hybrid population has been analyzed previously using interval mapping and linkage analysis.²⁷ This time, we employed single marker regression analysis using QTX Map Manager.⁵⁸

The algorithm of the third approach (results presented in Figure 3) was a modification of a single marker regression in QTX,⁵⁸ when mice were not grouped by congenic strains. Instead, we assumed that all mice, including WT BALB/c, belonged to a single advanced-backcross-like population. For each genomic marker, each mouse was considered to be either BALB/c-homozygous or DBA/2-homozygous at the given marker position, and the difference between the two groups for each clinical and immunologic trait was estimated using the Mann– Whitney U-test for two independent variables. A similar approach was successfully used for fine mapping of diabetes loci in a set of murine congenic strains.⁵⁹ Statistical significance for the association studies was expressed as a negative logarithm of P-values.

Statistical analysis

Statistical analysis was performed using SPSS software (SPSS, Chicago, IL, USA). Since the incidence of disease demonstrated a non-parametric distribution, we used the Mann– Whitney U-test to examine differences between populations. Calculations of the difference between a group of immunized WT and a group of immunized congenic mice were performed for every day of scoring. The two-sample non-paired Student’s t-test was employed for a comparison of the means of two groups where the data (arthritis score, severity, cytokine concentrations and stimulation indices) showed normal distribution. Analysis of variance was used for multiple comparisons. Additional information of statistical methods is incorporated in the legends of Tables 1 and 2. The significance level was routinely set at P<0.05.