Identification of arthritis-related gene clusters by microarray analysis of two independent mouse models for rheumatoid arthritis

Abstract

Rheumatoid arthritis (RA) is an autoimmune disease affecting approximately 1% of the population worldwide. Previously, we showed that human T-cell leukemia virus type I-transgenic mice and interleukin-1 receptor antagonist-knockout mice develop autoimmunity and joint-specific inflammation that resembles human RA. To identify genes involved in the pathogenesis of arthritis, we analyzed the gene expression profiles of these animal models by using high-density oligonucleotide arrays. We found 1,467 genes that were differentially expressed from the normal control mice by greater than threefold in one of these animal models. The gene expression profiles of the two models correlated well. We extracted 554 genes whose expression significantly changed in both models, assuming that pathogenically important genes at the effector phase would change in both models. Then, each of these commonly changed genes was mapped into the whole genome in a scale of the 1-megabase pairs. We found that the transcriptome map of these genes did not distribute evenly on the chromosome but formed clusters. These identified gene clusters include the major histocompatibility complex class I and class II genes, complement genes, and chemokine genes, which are well known to be involved in the pathogenesis of RA at the effector phase. The activation of these gene clusters suggests that antigen presentation and lymphocyte chemotaxis are important for the development of arthritis. Moreover, by searching for such clusters, we could detect genes with marginal expression changes. These gene clusters include schlafen and membrane-spanning four-domains subfamily A genes whose function in arthritis has not yet been determined. Thus, by combining two etiologically different RA models, we succeeded in efficiently extracting genes functioning in the development of arthritis at the effector phase. Furthermore, we demonstrated that identification of gene clusters by transcriptome mapping is a useful way to find potentially pathogenic genes among genes whose expression change is only marginal.

Figure 1

The relationship of gene expression levels between human T-cell leukemia virus type I- transgenic (HTLV-I-Tg) and interleukin-1 receptor antagonist-knockout (IL-1Ra-KO) mice. (a) The reciprocal relationship between log-transformed (base 2) normalized gene expression levels for the two models was plotted. The correlation coefficient of gene expression between HTLV-I-Tg and IL-1Ra-KO mice is r = 0.77. (b) Commonly activated genes were extracted in different models using the SAM (significance analysis of microarrays) method; their relationship is shown. The correlation coefficient is r = 0.91.

Figure 2

Distribution of arthritis-related genes on the mouse genome. The numbers of significantly changed genes are indicated for each chromosome (closed bars). The density of significantly changed genes (Number of significantly changed genes/Total number of genes on the chromosome estimated from the data of Mouse Genome Search in The Bioinformatics Analytical Toolkit; see Materials and methods) is shown for each chromosome (open circles).

Figure 3

Transcriptome mapping of arthritis-related genes. Significantly changed genes were mapped in every 1-megabase (Mb) interval on each chromosome. The chromosome number is indicated in each panel; the peak number corresponds to the number given in Table 1.

Figure 4

Identification of the H-2 gene cluster as one of the significantly activated gene clusters. The maximal gene density was mapped to the chromosome (Chr.) 17, 33- to 35-megabase (Mb) locus (#1–3), which corresponds to the H-2 gene cluster. Fold changes of the significantly changed genes are shown in a region of 2-Mb window. Hierarchical clustering of these genes is visualized below. Each column represents an RNA preparation from a different mouse strain, and each row represents an individual gene. Red represents expression levels greater than the median, and green represents those less than the median. The expression scale is shown at the bottom. KS, interleukin-1 receptor antagonist-knockout mice; MHC, major histocompatibility complex; TS, human T-cell leukemia virus type I- transgenic mice; WT, wild-type mice.

Figure 5

Arthritis-related gene clusters on chromosome (Chr.) 11. The peak at the Chr. 11, 81- to 83-megabase (Mb) (#4–6) includes Ccl and Slfn family clusters. An expanded view of this region in 2-Mb window is shown. Hierarchical clustering patterns of the expression levels are shown below. Ccl and Slfn genes are clearly clustered in specific narrow loci and augmented in arthritis. CC, Cysteine-Cysteine type; KS, interleukin-1 receptor antagonist-knockout mice; TS, human T-cell leukemia virus type I- transgenic mice; WT, wild-type mice.

Figure 6

Arthritis-related gene clusters on chromosomes (Chr.) 6, 15, and 19. The peaks on the Chr. 6, 124-megabase (Mb) locus (#12) includes complement receptor genes and C-type lectin superfamily (Clecsf) genes, the Chr. 15, 79-Mb locus (#13) includes colony-stimulating factor 2 receptor beta (Csf2rb) genes, and the Chr. 19, 11-Mb locus (#14) includes membrane spanning four-domain, subfamily A (Ms4a) genes. Expanded views of these loci in 1-Mb scale and hierarchical clustering of the expression levels of these genes are shown.