Objective: To define genomic regions that link to rat arthritis and to determine the potential association with rheumatoid arthritis (RA) of the corresponding human genomic regions.
Methods: Advanced intercross lines (AIL) between arthritis susceptible DA rats and arthritis resistant PVG.1AV1 rats were injected with differently arthritogenic oils to achieve an experimental situation with substantial phenotypic variation in the rat study population. Genotyping of microsatellite markers was performed over genomic regions with documented impact on arthritis, located on rat chromosomes 4, 10 and 12. Linkage between genotypes and phenotypes were determined by R/quantitative trait loci (QTL). Potential association with RA of single nucleotide polymorphisms (SNPs) in homologous human chromosome regions was evaluated from public Wellcome Trust Case Control Consortium (WTCCC) data derived from 2000 cases and 3000 controls.
Results: A high frequency of arthritis (57%) was recorded in 422 rats injected with pristane. Maximum linkage to pristane-induced arthritis occurred less than 130 kb from the known genetic arthritis determinants Ncf1 and APLEC, demonstrating remarkable mapping precision. Five novel quantitative trait loci were mapped on rat chromosomes 4 and 10, with narrow confidence intervals. Some exerted sex-biased effects and some were linked to chronic arthritis. Human homologous genomic regions contain loci where multiple nearby SNPs associate nominally with RA (eg, at the genes encoding protein kinase Cα and interleukin 17 receptor α).
Conclusions: High-resolution mapping in AIL populations defines limited sets of candidate risk genes, some of which appear also to associate with RA and thus may give clues to evolutionarily conserved pathways that lead to arthritis.
Statistics from Altmetric.com
Chronic polyarthritis that fulfils the criteria for rheumatoid arthritis (RA)1 can be induced in genetically predisposed rat strains by a single intradermal injection of molecules that trigger the innate immune system,2 resulting in pathogenic T helper cells that exclusively target peripheral joints.3 Clues to the enigmatic disease pathways of this disease syndrome may come from the identification of underlying genetic determinants. Beside the major histocompatibility complex (MHC), only two novel risk genes or types of genes have so far been position mapped: the neutrophil cytosolic factor 1 (Ncf1),4 and the antigen-presenting lectin-like receptor gene complex (APLEC),5 which harbours immunoregulatory genes, including the dendritic cell immunoreceptor (DCIR).5 6 7 8 In RA, initial genetic association results are encouraging for APLEC and for Ncf4, which like Ncf1 is one of several components in the nicotinamide adenine dinucleotide phosphate (NADPH)–oxidase complex.9 Consequently, Ncf1 and APLEC/DCIR may both define disease pathways that are conserved through evolution and that can be manipulated to achieve disease prevention and treatment. However, a limiting factor in the comparative genetic approach is the time-consuming production of congenic strains for positional mapping. In the present study we describe results of an alternative method. We aimed to define candidate risk genes in regions on Rattus norvegicus chomosome 4 and 10 (RNO4 and 10) that influence experimental RA,10 11 12 13 14 15 16 and in the corresponding human chromosome regions which link to, and associate with, RA.17 18 19 As well as RA, the targeted regions also associate with other inflammatory diseases, and there is evidence for more than one risk gene in each region.20 21 22 23 24 25 26 27 28 29 30 31 32 33
First, we performed linkage mapping in advanced intercross lines (AIL), which is a continuously and randomly intercrossed rat population that allows for simultaneous identification of genetic interactions and high-resolution mapping of multiple quantitative trait loci (QTLs).34 A seventh generation (G7) AIL was established by interbreeding arthritis-prone DA rats with arthritis resistant but MHC identical PVG.1AV1 rats, to achieve phenotype differences in the progeny exclusively controlled by non-MHC genes. Second, AIL study populations were exposed to oils with differential arthritogenic capacities. This “genome environment” approach35 36 identified pristane as an effective trigger of arthritis and chronic disease. Third, the pristane-exposed rat population was genotyped with genetic markers covering the targeted RNO4 to RNO10 regions. As a control for mapping precision and resolution, the RNO12 region harbouring Ncf1 was also evaluated. Fourth, the selected chromosomal intervals were scanned for QTLs underlying associations between phenotypes and genotypes, and for potential genetic interactions between QTLs. Finally, public databases were used to translate rat QTLs into homologous human chromosomal regions and to retrieve applicable RA association data from 2000 cases and 3000 controls, as generated by the Wellcome Trust Case Control Consortium (WTCCC).37
Materials and methods
The AIL rats were produced by an initial intercross between DA and PVG.1AV1 rats. The F1 population originated from two breeding pairs with DA and PVG.1AV1 female founders, respectively. F2 was generated from 14 breeding pairs, 7 F1 couples each from the DA and PVG.1AV1 female founder populations, respectively. The G3 generation originated from a breeding F2 population of 100 rats, with both types of female founders, and this was repeated for subsequent generations avoiding brother–sister mating by controlled breeding. Rats were bred and kept at the Center for Molecular Medicine (Karolinska Institutet, Stockholm, Sweden) in a 12h light/dark cycle, housed in polystyrene cages containing Aspen wood shavings with free access to water and standard rodent chow. They were routinely tested negative for specific pathogens according to a health-monitoring programme for rats set up at the National Veterinary Institute in Uppsala, Sweden. The local ethical committee in Northern Stockholm approved the experiments.
Arthritis induction and evaluation
Animals aged 120–160 days received one intradermal injection at the dorsal base of the tail with one of the following agents: 200 μl squalene (C30H50; Sigma, St Louis, Missouri, USA), 200 μl incomplete Freund adjuvant (IFA), or 150 μl pristane (C19H40; Sigma). The animals were weighed and visually examined every second day, and arthritis was scored for each paw as follows: 0, no joints affected; 1, one type of joint affected (redness and/or swelling): 2, two types of joints affected; 3, three types of joints affected; and 4, the entire paw affected. The types of joints examined were peritarsal, intratarsal and ankle joints. Scores were added, yielding a total score for all four limbs ranging from 0 to 16. Animals were scored until 70 days postinjection. Signs of chronic arthritis were evaluated from 40 days postinjection and every 10 days by comparing paw maps depicturing inflamed joints at a given timepoint. Arthritis was classified as chronic when new sites of inflammation appeared over time, as previously described.1 Other macroscopic phenotypes analysed were: incidence, the cumulated number of affected rats (scores⩾1) over time, divided by all tested rats; severity, the maximal score (1–16) attained by each affected rat (scores 1–16); Day of onset, the first recorded sign of joint inflammation; and sensitivity, the maximal score (0–16) attained by each tested animal.
DNA was purified from individual tail tips38 and subjected to genotyping with 54 microsatellite markers (see Supplementary material). The polymerase chain reaction (PCR) was performed as previously described,39 except the primers were labelled with [γ-33P]ATP. The polymorphic PCR products were size separated by electrophoresis on 6% polyacrylamide gels and later visualised by autoradiography. Marker information was retrieved from the rat genome database (http://rgd.mcw.edu/), except for microsatellite sequences that were retrieved from the National Center for Biotechnology Information (NCBI) rat genome sequence database (http://www.ncbi.nlm.nih.gov) in order to design new primers using the Primer 3 software (http://frodo.wi.mit.edu/primer3/primer3_code.html).
Linkage analyses and statistical analyses
Genotyping success rates were high (average: 99.45%, range 97.0% to 100%), recombinations occurred between all adjacent microsatellites (0.24% to 25.6%) and no marker homozygosation was observed. Linkage analysis was performed using R/QTL software.40 41 The physical map is derived from the NCBI rat genome sequence at http://www.ncbi.nlm.nih.gov. Marker positions were obtained from the NCBI rat genome assembly 2. Marker order was confirmed by building a genetic map in MAPMAKER (http://www.broadinstitute.org/ftp/distribution/software/mapmaker3) using genotype data from the G7 population. Data were analysed implementing non-parametric model for quantitative traits and binary model for categorical traits. Confidence intervals (CIs) for QTLs were defined as the region within maximum logarithm of odds (LOD) minus 1 LOD. To further evaluate identified QTLs, a multiple QTL model test was performed using R/QTL software. This analysis is based on the creation of an initial model of phenotypic variance, comprising all identified QTLs displaying significant linkage. Each QTL is then excluded from the initial model in subsequent steps and the influence on the phenotypic variance by the excluded QTL is determined. The outcome of the multiple QTL model analysis provides information on whether a QTL has a significant effect on the phenotype (p⩽0.05) or not (p>0.05). Epistatic interactions and additive effects were searched for using a two-dimensional scan with a two-QTL model. Mann–Whitney U test was used to test for differences in sensitivity, severity and day of onset. A χ2 test with 1 degree of freedom was used to test for differences in incidence and chronicity. Tests were performed using Statview 5.0 (Abacus Concepts, Berkeley, California, USA). p Values below 0.05 were considered significant.
Evaluation of three oil-induced arthritis models in G7 (DA × PVG.1AV1) populations
Following challenge with IFA, squalene, or pristane, we recorded the development of oil-induced arthritis (OIA),42 squalene-induced arthritis (SIA),1 2 and pristane-induced arthritis (PIA),43 respectively (see Supplementary material). Inflammation affecting peripheral joints, and with similar severity, developed in all 3 populations, but the incidence was low in OIA (8%, 4/50), medium in SIA (25%, 27/107) and high in PIA (58%, 33/57). Chronic arthritis developed in all populations, but with high frequency only in PIA (30%, 17/57), where it occurred in the majority of affected individuals (58%, 17/33). Based on these results, pristane was used to challenge two additional populations of G7 AIL rats (161+204 rats), yielding similar results as in the first set (57 rats) and summarised in table 1.
In the total number of 422 pristane-injected rats the following phenotypes were recorded: incidence 57% (240/422); severity 6.1 (3.6); sensitivity 3.5 (4.1); mean day of onset 23 (14); and chronic arthritis 21% (87/422). Females generally developed more pronounced phenotypes than males, with significant differences for incidence (64% vs 48%, p<0.001) and sensitivity (4.0 (4.2) vs 2.8 (3.8), p<0.001). All pristane-injected rats were subsequently genotyped.
Fine mapping of the RNO12 region harbouring Ncf1
A 4.2 mega-base pair (Mb) RNO12 region harbouring Ncf1 was genotyped with nine microsatellite markers. Subsequent linkage analysis suggested an influence from Ncf1 on all arthritis phenotypes, with sensitivity and incidence surpassing the LOD threshold for significance (fig 1, tables 2 and 3). The RNO12 CI for sensitivity spanned 1.1 Mb (22.9 to 24.0 Mb), with the maximum LOD score at 23.47 Mb, only 100 kb from D12Kir7. This marker is located between exon 7 and 8 in the Ncf1 gene, which corresponds to Pia4.4 Since PVG rats carry the Ncf1 resistance allele,44 while DA rats do not, our data almost certainly reflect Ncf1 impact.
Fine mapping of the telomeric end of RNO10
The 19 Mb telomeric end of RNO10 was genotyped with 23 microsatellite markers. Subsequent linkage analysis identified two arthritis-linked QTLs (fig 1, tables 2 and 3). One locus influencing sensitivity was defined by D10Rat13, at 97.2 Mb, near Prkca, with a 1.7 Mb CI (96.7 to 98.4 Mb). Another locus influencing sensitivity and incidence was defined by D10Got158, at 105.1 Mb, near Cd300le, with a 3.5 Mb CI (104.5 to 108.0 Mb). The two QTLs were verified when subjected to a multiple QTL model test (table 2) and were designated Pia30 and Pia31, respectively.
Fine mapping of the RNO4 region harbouring APLEC
A 17 Mb RNO4 region harbouring APLEC was genotyped with 22 markers. Subsequent linkage analysis did not reveal a single sharp LOD peak as for Ncf1. Rather, several QTLs appear to link to various arthritis phenotypes, and often with sex-specific effects (fig 1, tables 2 and 4). One QTL designated Pia27, with a 2.6 Mb CI (157.5 to 160.1 Mb), regulates sensitivity, incidence and chronicity. Pia27 appears to harbour two sub-QTLs, although this could not be formally verified in a multiple QTL model test (table 2). One potential sub-QTL, Pia27a, was defined by the marker D4Got126, located at 159.33 Mb, only 130 kb centromeric of APLEC, which corresponds to Oia2.5 APLEC/Oia2 was mapped in OIA using congenic strains derived from the same two strains used here to generate the AIL (ie, DA and PVG.1AV1). Therefore, the present maximum linkage near APLEC/Oia2 strongly suggests that Pia27a contains, or is identical to, Oia2/APLEC. A second potential sub-QTL, Pia27b, was defined by D4Rat62 at 157.9 Mb, close to A2m. Interestingly, Pia27 is markedly linked to chronicity, with higher LOD scores when analysing females separately, despite loss of statistical power due to decreased number of animals. Chronicity in females appears also to be regulated by a novel definite QTL designated Pia28, located at 161.0 Mb close to Cd4, with D4Got34 as the nearest marker and with a 5.5 Mb CI (160.6 to 166.1 Mb). A third definite QTL influencing severity in males, designated Pia29, is located at 156.5 Mb, defined by D4Rat63, near the Ninj2 gene, with a 2.2 Mb CI (155.1 to 157.5 Mb).
Additive effects and epistatic interactions
Analyses of the genotype and phenotype data with a combinatory two-QTL model test (table 5) revealed that each QTL exerts additive effects with one or several other QTLs (eg, concerning chronicity (threshold 6.5): RNO12 (24 Mb) with RNO4 (158 Mb) LOD15.4, RNO12 (24 Mb) with RNO10 (97 Mb) LOD 11.80, and RNO10 (105 Mb) and RNO4 (158 Mb) LOD 11.23). The analysis also suggested two novel and epistatically-interacting RNO10 QTLs that influence chronicity only in DA allele combinations, LOD 5.03 (threshold 4.98). These two QTLs, at 100 (0.5) Mb and 104 (0.5) Mb, were undetectable in the one-dimensional linkage analysis on RNO10, as depicted in fig 1.
Association with RA of candidate risk genes derived from the rat AIL mapping
Candidate risk QTLs in Homo sapiens (HSA) were defined by homology comparisons (http://www.ensembl.org) as follows: Pia29+Pia27+Pia28 is approximate to HSA22 (15.85 to 16.75 Mb)+HSA12 (0 to 9.75 Mb); Pia30 is approximate to HSA17 (60.95 to 63.10 Mb) and Pia31 is approximate to HSA17 (69.7 to 74.0 Mb). Summary genotype data (general model) for association of SNPs with RA were retrieved from WTCCC37 (http://www.wtccc.org.uk). While this approach does not consider underlying genomic structure, it can point to candidate risk genes for focused studies. No associations were significant following conservative Bonferroni corrections. Notably large regions even lacked occasional SNPs with even nominal association (p<0.05), while several loci were characterised by multiple such SNPs, as would be expected for most real susceptibility loci37 (see Supplementary material). For example, the screened 2.15 Mb HSA17q24 (60.95 to 63.10 Mb) region contains data for 363 SNPs. Of these, 20 SNPs show nominal association (0.05<p<0.001). They are all located within a 1.31 Mb subregion covered by 231 SNPs (spanning the gene sequence PRKCA–CACNG5–CACNG4–CACNG1–PSMD12–HELZ–PITPNC1 and with most of the 20 SNPs within or adjacent to PRKCA. Notably, most of these SNPs show differential effects in males and females (0.05<p<0.001, see Supplementary material), just like rat Pia30, where maximum linkage occurs at Cacng4. Notably, the 20 SNPs are surrounded by >2.1 Mb with >284 SNPs, none of which associate with arthritis. Similar concentration of putatively associated SNPs were observed on HSA22 (15.85 to 16.75 Mb), at the IL17RA–CECR6–CECR5–CECR1 locus (see Supplementary material).
Our rat linkage-mapping data reveals an extraordinary resolution and precision, with a maximum LOD score at the Pia4 RNO12 locus Ncf1, and a narrow 1.1 Mb CI containing only 16 genes or predicted genes. This mapping resolution is high compared to previous results in arthritis using AIL, or partial AIL,45 46 47 48 49 and it approaches the mapping results for Ncf1 and APLEC in congenic strains (<0.6Mb).
Our accurate pinpointing of the Ncf1 reference gene suggests that other QTLs may be mapped with similar precision in the AIL population. In fact, the Oia2 RNO4 locus APLEC, here corresponding to Pia27a, was also precisely mapped, with maximum LOD only 130 kb to the centromeric side. But while Oia2 was originally mapped in monophasic OIA, we demonstrate here that Pia27a also links to chronic arthritis in females. Interestingly, this is an effect we have hitherto not documented in PIA, or any other form of arthritis, using a congenic strain with DA background and PVG alleles in Oia2/APLEC.5 6 This supports our present mapping results demonstrating arthritis-regulating loci nearby APLEC in the screened RNO4 interval.
Although we did not monitor bone erosion, our present demonstration of loci mediating female biased effects on chronicity is of particular interest for RA where female preponderance is a disease hallmark. We suggest continuing studies on the corresponding human 12p13 region in arthritis and inflammatory diseases, focusing on APLEC and DCIR5 and taking genetic interactions and additive effects into account. Such cumulative effects may explain previous linkage results to the region, and also why no single associations withstood Bonferroni correction in this study, where all human data should be cautiously interpreted until replicated.
On RNO10, we mapped two novel QTLs, Pia30 and Pia31, with 1.7 to 3.5 Mb resolution, compared to 10.1 to 47.3 Mb in recently published mapping papers.15 50 Beside these novel QTLs identified by one-dimensional linkage analysis, some evidence for two intervening QTLs were obtained following a two-dimensional scan that unmasked loci involved in epistatic interactions. Consequently, association testing of the corresponding human candidate disease genes should take epistatic interactions into account. Here, we took the opportunity presented by the WTCCC to scan the regions of interest for evidence of association with RA. One region that appears to associate with RA contains the following genes on 17q24: PRKCA–CACNG5–CACNG4–CACNG1–PSMD12–HELZ–PITPNC1. Interestingly, linkage to, and association with, this region has previously been reported in RA families.17 Furthermore, exactly the same genes were highlighted in a recent association study in multiple sclerosis, with strongest evidence for PRKCA in Canadian and Finnish populations,51 and with similar data in a UK population as well.52 PRKCA is an excellent functional candidate for predisposition to RA, because it is involved in Ca2+-induced regulation of a diverse set of cellular responses, such as ion channel activation, apoptosis, proliferation and G-coupled regulation of inflammatory cells. PRKCA is expressed in many cell types, and, for example, plays a critical role in signal transduction controlling T cell activation.53 54 From our comparative genetics study, we forward PRKCA as one example of a highly relevant candidate risk gene. Other candidates for focused follow-up studies include IL17RA and nearby genes on 22q11.2, as well as a non-APLEC gene cluster encoding C-type lectin domain receptors on 12p13, which was pointed out in an association study of type 1 diabetes by the WTCCC.37
We thank Holger Luthman for critically reading the manuscript.
▸ Additional data (Supplementary tables 1–6) are published online only at http://ard.bmj.com/content/vol68/issue12
Funding This work was supported by the Swedish Research Council, the Swedish Rheumatism Association, King Gustaf V 80th Birthday Foundation, Åke Wibergs Foundation, Alex and Eva Wallström Foundation, Professor Nanna Svartz’ Foundation and the FP6 EU-project AutoCure LSHB-CT-2006-018661. Pfizer contributed to the RNO4 analyses.
Competing interests None.
Ethics approval The local ethics committee in Northern Stockholm approved the experiments.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.