Comparative analysis between Rheumatoid arthritis and arthritis model – study of the functional components in expression profiles of synovitis Irene Ziska 1, Marc Bonin 1, Jekaterina Kokatjuhha 1, Pascal Schendel 1, Karsten Mans 1, Biljana Smiljanovic 1, Till Sörensen 1, Andreas Grützkau 2, Bruno Stuhlmüller 1, Thomas Häupl 1 1 Department of Rheumatology and Clinical Immunology, Charité University Hospital, Berlin, Germany, 2 German Arthritis Research Center, Berlin, Germany Acknowledgement: BTCure IMI grant agreement no ArthroMark grant no 01EC1009A Background and Objective: Materials and Methods: Results: Conclusion: Transcriptomes of immune cell types including T-cells, B-cells, NK-Cells, monocytes and granulocytes as well as transcriptomes of synovial tissue biopsies from patients with rheumatoid arthritis and osteoarthritis were generated using the Affymetrix platform HG-U133Plus 2.0. The MG-4302 GeneChip transcriptomes of the same panel of cells (GSE6506) and synovial tissues from the collagen induced arthritis model in DBA-1 mice (GSE13071) were selected from the GEO database. We developed a model to compute cell fractions for each sample by using cell-specific marker genes from reference transcriptomes for each cell type involved. (Fig. 1) Based on these fractions a virtual profile was computed for each sample that represents the sample-specific mixed cell profile based on the reference transcriptomes and thus without the disease-related regulation of genes. These so called virtual signals were compared to the corresponding real signal to estimated the level of gene regulation. The given model was used to compare the gene regulation in the human disease of rheumatoid arthritis with the collagen induced arthritis model in DBA-1 mice. This new analysis technique revealed activity dependent infiltration of monocytes into the synovial tissue in the DBA-1 mouse model. In contrast to the mouse model, the human disease transcriptomes included patterns of T-cells and B-cells as well as monocytes. With respect to the regulatory changes, there were significant differences not only by quantity but also by quality. For example differences especially in chemokine regulation like CXCL13 between human disease and animal model are in accordance with the observed difference in cell type infiltration. (Fig. 2) In summary a model to differentiate between a change in expression level based on regulation and a change due to a different cell composition was developed and successfully applied for synovitis. The data indicate differences between human disease and animal model and can provide important information on the selection of appropriate animal model for the development of therapeutic targets in human diseases. Microarray experiments can be used to compare samples of healthy and diseased patients. The objective is to identify up or down regulated genes in the diseased sample which could be the key to understand the disease and develop a treatment. The Problem is that the taken samples often not only consist of one but of multiple different cell types and this composition can greatly differ between the healthy and the diseased sample because of the migration of immune cells. Therefore it is important to distinguish if an expression level of a gene is changed due to regulation or as a result of a change in the cell composition. (Fig. 3) Figure 1: In the figures the cell fractions are shown (human l., mouse r.) that were calculated by the given model. The mean values for each group are highlighted. In both human and in the mouse model the cell fractions of synovial tissue are reduced in the diseased probes. In contrast the cell fractions of immune cells are increased. Monocytes have the greatest cell fraction compared to the other immune cells. In human for all immune cells an increase of the cell fraction can be observed. But in the mouse model only the fractions of monocytes, T-cells and NK cells have changed. Therefore the greatest difference between human disease and animal model is the unchanged fraction of the B-cells in the mouse model because the B-cells play an important role in rheumatoid arthritis. All in all there are differences between the cell compositions of healthy and diseased probes and differences between human and mouse model that could be very important for drug designing and testing therapies in the given mouse model. Figure 2: The virtual signals that were calculated by the model were used to find genes that have a change in their signal due to a regulation and not only because the cell composition has changed. Afterwards the real signals of these genes were clustered and compared to the signatures of TNF stimulated synovial fibroblasts. In both human (l.) and the mouse model (r.) similarities between the up regulated genes in RA and accordingly in the arthritis model and the TNF stimulation are a sign of a good filtering because anti-TNF is already used in the therapy of rheumatoid arthritis. Additionally the similarities with the signatures of the immune cells are quite low. Therefore in the list of genes that was the result of the filtering no genes were included whose signal values were only changed due to migrated immune cells. That is a fact which supports the correct function of the developed model. Marc Bonin Department of Rheumatology and Clinical Immunology Charité University Hospital Charitéplatz 1 D Berlin Germany Tel: +49(0) Fax: +49(0) Web: Contacts: Figure 1 Figure 2 Figure 3 Figure 3a: Gene regulation vs. change of the cell composition. Figure 3b: Shown is the measured signal of a gene for OA and RA. Without calculating the FPCA it is not possible to know the cell composition. After the FPCA, it is clear that the signal in the RA is strongly influenced by monocytes and the synovial tissue which have migrated into the tissue. regulation change in the cell composition a b