Materials and methods

Animals, immunisation and assessment of arthritis
Both DBA/1 and FVB/N mice used in this study were obtained from the Jackson Laboratory and kept in a climate-controlled environment with 12 hour light/dark cycles in the animal facility at the University of Rostock. All animal experiments were pre-approved by the State Animal Care Committee. CIA was induced in control and experimental animals according to established protocols described previously [10]. In brief, DBA/1J, FVB/N and (DBA/1J × FVB/N)F2 progeny were immunised at 8 to 12 weeks at the base of the tail with 125 μg of bovine Collagen II (Chondrex, Redmond, WA, USA) emulsified in CFA (DIFCO, Detroit, MI, USA). The clinical scoring of arthritis commenced 18 days after immunisation, and animals were monitored three times weekly for signs of CIA. Arthritis development was monitored in all four limbs using a three-score system per limb as described previously [10].
Eight-week old FVB/N and DBA/1J mice were used for the detection of gene expression. They were divided into four experimental groups according to the different phases of CIA, namely naive control (NC), post-immunisation (PI), onset of arthritis (OA) and chronic arthritis (CA) (Table 1). The NC group contained non-immunised mice that were sacrificed at the age of 8 weeks. The mice in the PI group were sacrificed on day 10 after immunisation. The mice in the OA group were sacrificed on day 35 after immunisation. Three FVB/N non-arthritic mice and three DBA/1 mice that showed signs of arthritis on day 33 or 34 after immunisation were sacrificed on day 35 after immunisation. The mice in the CA group were sacrificed on day 95 after immunisation. Three non-arthritic FVB/N mice and three DBA/1 mice that had developed arthritis for at least two months were sacrificed on day 95 after immunisation.

Linkage analysis
Detailed information on genotyping of the genome screen has been described previously [10]. In short, we genotyped 290 F2 mice using 126 informative microsatellite markers covering the genome with an average inter-marker distance of 11.5 cM for 290 F2 progeny. All linkage analyses were performed with QTX Map manager software [23]. The main clinical phenotype of CIA, arthritis severity, was taken as phenotype. To detect the small-effect QTL, the threshold value of linkage was set as P = 0.05 (Chi-square test).

Sample preparation and microarray hybridisation
Lymph nodes (LNs) draining the immunisation site were used for total RNA preparation. The total RNA was extracted from the tissue homogenates using a commercial kit in accordance with the provided protocol (QIGEN, Hilden, Germany). Analysis of gene expression was conducted using a U430A array (Affymetrix, Santa Clara, CA, USA) interrogating more than 22,000 genes. RNA probes were labelled in accordance with the manufacturer's instructions. Samples from individual mice were hybridised onto individual arrays. Hybridisation and washing of gene chips were done as previously described [16]. Fluorescent signals were collected by laser scan (Hewlett-Packard Gene Scanner).

Microarray analysis
Normalisation of the expression level was done using Affymetrix software MAS 5, which is based on global scaling of total gene expression level per microarray. The normalised expression values were imported to and analysed by dCHIP [24]. Differentially expressed genes were identified by defining the appropriate filtering criteria in the dCHIP software as: lower 90% confidence boundary of fold change between the group means exceeded twofold; the absolute difference between the two groups exceeded 100; the P value threshold of the unpaired t-test was 0.05. The false discovery rate was established with a permutation test for each pairwise comparison to estimate the proportion of false-positive genes.
Hierarchical gene clustering was performed with dCHIP to characterise the gene expression patterns during CIA. The default clustering algorithm of genes was as follows: the distance between two genes is defined as 1 – r, where r is the Pearson correlation coefficient between the standardised expression values of the two genes across the samples used. To characterise the functional relationship between differentially expressed genes, Gene Ontology (GO) term classification incorporated in DNA-Chip Analyzer was performed. The significant level for a function cluster was set at P < 0.005, and the minimum size of a cluster was three genes.