Microarray data analysis

Hierarchical clustering

This gene list was subjected to a hierarchical clustering routine in GeneSpring using a Pearson correlation similarity measure across conditions and a standard correlation similarity measure across genes. The final tree is represented in Figure 1.

Pathway Analysis

The 1.5 fold list was further analyzed using Onto-Express and Pathway-Express, part of the Onto-Tools package available from the Draghici laboratory. This software allows for identification of pathways that show enrichment in the microarray data and is a novel tool that translates lists of differentially regulated genes into functional profiles characterizing the impact of the condition studied.

Quantitative PCR validation

To verify expression data obtained by the Codelink analysis, Taqman gene expression assays were obtained (Applied Biosystems) to evaluate the expression of selected genes by RT-qPCR. The assays utilized were as follows: TCF2 (Applied Biosystems assay Hs01001609_m1), CXCL10 (Hs00171042_m1), FUT4 (Hs00275643_s1), SERPINB1 (Hs00229084_m1), CD9 (Hs01124026_m1), LYPLA3 (Hs00202399_m1), SERPINB9 (Hs00394497_m1), CD83 (Hs00188486_m1), CDC5L (Hs00904622_m1), TNFRSF19 (Hs00218634_m1), HLA-DRA (Hs00219578_m1). All assays are designed to detect the appropriate transcripts at an exon-exon junction to ensure the detection of mRNA and not genomic DNA.

The expression levels of these 11 genes were evaluated by two-step Taqman-based RT-qPCR. First strand cDNA was synthesized with M-MLV reverse transcriptase (Ambion, Inc., Austin, TX) and primed with oligo-dT per the manufacturer's specifications. The cDNA equivalent of 20 ng starting RNA was then included in reactions containing 1× Taqman Master Mix (Applied Biosystems, Foster City, CA), and 1× gene-specific assay reagents as recommended by the manufacturer (Applied Biosystems). All reactions were run in triplicate. Reactions that did not contain template cDNA were included as negative controls.

Reaction plates were processed on an Applied Biosystems 7900 HT Sequence Detection System. The AmpliTaq Gold polymerase was activated at 95°C for 10 min followed by 40 cycles consisting of denaturation for 15 seconds at 95°C and annealing and extension for 60 seconds at 60°C.

Amplification data was analyzed with the ABI Prism SDS 2.1 software (Applied Biosystems). Although run individually, samples GM8921A, 9017, 9018, and 9023 were pooled for analysis as the control sample, and samples GM8830, 8922A, 8923, 8839, 9013, and 9016 were pooled for analysis as the MS sample, unless otherwise noted. Relative quantification of gene expression was performed by the ΔΔCt method [18]. GAPDH expression was used as an endogenous control to normalize expression within each sample.

Transcription network analysis

We employed a software program that one of our authors continues to develop (G.A.) to recast the matrix of regulatory interactions found in a prior step as a directed graph overlaying color information for gene responses. Our regulatory network was constructed by integrating transcription factor binding site information (from TRANSFAC) with gene expression data using graph TFs (Bioconductor functions for transcription network analysis). The standard microarray experiments cannot measure the transcription factor activities (TFAs) directly, since TFAs are subject to post-translational modifications or ligand binding in order to exert their function.

We hypothesized that decoding some of the important TFAs involved in B cells of MS patients might provide important clues to disease pathogenesis or uncover potential molecular targets for therapy. We started with an initial list of all genes changing at the 1.5-fold level. In this dual sample study, we assumed all pairs of genes moving in the same direction to be concordant and all pairs moving in opposite directions to be discordant. Combining this with the potential for regulatory relationships inferred from a TRANSFAC binding site dataset, we were able to extrapolate the potential transcription regulatory networks driving expression changes. We found that TCF2 and CDC5L bind very similar promoter motifs and seem to control overlapping sets of up and down regulated genes in this study. The transcripts we uncovered in this dataset are most notably, CCR1 and CXCR4, molecules which are associated with MS. The multiplication of corresponding cells in the TFBS and correlation matrices yields a regulation matrix where the sign of the interaction score reflects positive or negative regulation and the magnitude of the score reflects the amount of supporting evidence, as shown in the grid below (Figure 2). We examined these networks for various thresholds of fold-change and TRANSFAC evidence scores to obtain the picture of regulation of downstream molecules by TCF2 and CDCL5 shown in Figure 3.

Additional data

A list of genes at 1.5 fold differential expression from microarray data can be found in Additional file 1.

Study limitations

The findings in our study are limited by a) sample size and b) the cellular pathways that Epstein-Barr virus (EBV) regulates remain poorly characterized and there are few data on how EBV may influence gene pathways in B cells derived from MS patients vs. controls. The ideal starting material would be to obtain native B cells from controls and MS patients; however, collection and processing bias are hard to eliminate in any microarray-based study.

Upregulated genes

The gene for CXCL10 (interferon-γ-inducible protein-10) encodes for an interferon (gamma)-induced, secreted protein of 10 kDa, a chemokine of the CXC subfamily that is one of the ligands for the receptor CXCR3. The binding of this protein to CXCR3 causes pleiotropic effects, including stimulation of monocytes, natural killer and T-cell migration, and modulation of adhesion molecule expression (UCSC web browser). Its levels are increased in cerebrospinal fluid of MS patients with symptomatic attacks of demyelination, suggesting a role for this molecule in MS [21]. Following mouse hepatitis virus (MHV) infection in mice [22], CXCL10 (IFN inducible protein 10 kDa) protein was expressed during both acute and chronic stages of disease suggesting a role for this protein in disease exacerbation and chronicity. Previous studies have shown that during the acute phase of infection, T lymphocytes are recruited into the CNS by CXCL10. Our finding that CXCL10 upregulation (5.8-fold) in B cells from MS siblings is supportive of prior observations.

Serpin B1 is a leukocyte elastase inhibitor and regulates the activity of the neutrophil proteases elastase, cathepsin G and proteinase-3. In humans, serpins constitute 10% of the plasma proteins and are known as critical regulators of both thrombotic and fibrinolytic systems. Serpins participate in the regulation of the complement cascade, angiogenesis, apoptosis and innate immunity. Most of the human clade B serpins inhibit serine and/or papain-like cysteine proteinases and protect cells from exogenous and endogenous proteinase-mediated injury. As some serpins also guard cells against the deleterious effects of promiscuous proteolytic activity [23], it is possible that the cytoprotective function is a common feature of intracellular serpin clades and we hypothesize that up-regulation of serpin B1 gene may be protective in MS. While our findings need to be validated, serpin B1 gene may represent a novel therapeutic target to ameliorate MS, considering the importance of these molecules in regulating proteolytic cascades.

We found that FUT4 gene was upregulated by 2.5 fold. It is a member of the interleukin 1 cytokine family and the protein encoded by the gene is produced by activated macrophages as a pro-protein and processed to its active form by caspase 1 (CASP1/ICE). This cytokine is an important mediator of the inflammatory response, cell proliferation, differentiation, and apoptosis. The induction of cyclooxygenase-2 (PTGS2/COX2) by this cytokine in the central nervous system (CNS) is found to contribute to inflammatory pain hypersensitivity. FUT4 mRNA is increased in Jurkat cells undergoing apoptosis.

LYPLA3 (lysosomal phospholipase A2): Lysophospholipases are enzymes that act on biological membranes to regulate the multifunctional lysophospholipids. The protein encoded by this gene hydrolyzes lysophosphatidylcholine to glycerophosphorylcholine and a free fatty acid. LYPLA3 is present in the plasma and thought to be associated with high-density lipoprotein. Cellular phospholipases are key participants in cellular transduction [24] and are thought to be involved in the pathogenesis of local and systemic inflammatory disorders [25]. Interestingly, cytosolic phospholipase A2 levels were found to be low, possibly reflecting proteolysis or inactivation of enzyme activity in brain MS lesions [26].

The gene for XCL1 was upregulated; it encodes for chemokines, a group of small (8-14 kD) molecules that regulate cell trafficking of leukocytes through interactions with a subset of 7-transmembrane, G protein-coupled receptors. Chemokines also play fundamental roles in the development, homeostasis, and function of the immune system, and they have effects on cells of the central nervous system as well as on endothelial cells involved in angiogenesis or angiostasis. The protein product of XCL1 has chemotactic activity for neutrophils may play a role in inflammation and exerts its effects on endothelial cells in an autocrine fashion. We also found PSTPIP1 (proline-serine-threonine phosphatase interacting protein1) upregulated; its role in MS is unclear.

Downregulated genes

An important finding in our study is an 11.7-fold downregulation of eukaryotic translation initiation factor 1A, Y-linked (EIF1AY) gene. It is thought that EIF1AY encodes for minor histocompatibility antigen (mHA) and B-cell mediated antibody response to Y-chromosome encoded histocompatibility antigens (H-Y antigens) is associated with maintenance of disease remission in graft-vs-host disease [27]. The B cells may provide a new target for immune intervention in chronic graft vs. host disease and whether such a correlation can be translated to chronic MS requires further study.

We found that the tumor necrosis factor receptor (TNFR) superfamily 19 gene (TNFRSF 19) was downregulated by 5.3-fold. In a recent study by Achiron et al., TNF superfamily members 14, 1B, 4, and 6 were downregulated. The TNFR superfamily of cell surface receptors share approximately 80 amino acids within their cytoplasmic region called the death domain (DD), critical for recruiting the death machinery [28]. Individual TNF receptors are expressed in different cell types and have a range of affinities for various intracellular molecules that provide signaling and biological specificities. However, much is unknown about how individual signal transduction pathways are regulated upon activation by any particular TNF receptor, under physiological conditions [29] and while TNFR superfamily 19 genes may mediate important functions in immunity, inflammation, differentiation, cell proliferation control, and apoptosis, their specific effects in MS are unknown.

The gene for CD83 was downregulated in MS siblings. CD83 molecules are expressed at a high level on immune-competent, activated and mature dendritic cells. In MS, CD83-positive dendritic cells are present in occasional perivascular cuffs in lesions, suggesting that mature dendritic cells migrate to the CNS in response to chemokine signals [30].

The downregulation of insulin receptor tyrosine kinase substrate (BAIAP2L1) gene in B cells of MS siblings is significant. The insulin-like growth factor I receptor (IGF-IR), is a member of the receptor tyrosine kinase family of growth factor receptors. In rheumatoid arthritis (RA) it has been shown that IgG antibodies from the sera of patients can stimulate synovial fibroblasts through interaction with insulin-like growth factor receptor 1 (IGF-R1), provoking trafficking of T cells [31]. The data demonstrate, for the first time, a bridging link between B-cell activity and T-cell trafficking. In addition, they are of potential importance for the development of innovative therapeutic strategies, in which interrupting the IGF-1/IGF-1R axis could result in sustained disease modification by affecting both the growth-factor triggered activation of fibroblasts and the accumulation of T lymphocytes. A similar link in MS has not been demonstrated but is plausible. Other genes downregulated include ADAMTS16, BID, MIF and DAPK2, involved in the apoptosis pathway.

TCF2, diabetes and MS

Our transcription and microarray analysis showed that TCF2 was upregulated. TCF2 encodes transcription factor 2, a liver-specific factor of the homeobox-containing basic helix-turn-helix family. The TCF2 protein forms heterodimers with another liver-specific member of this transcription factor family, TCF1; depending on the TCF2 isoform, activation or inhibition of transcription of target genes occurs. We found that TCF2 upregulates the CC chemokine receptor 1 (CCR1) a molecule expressed on macrophages and T cells; in MS plaques, numerous CCR1-positive infiltrating macrophages and microglial cells, associated with CCL3, were found [2] and CCR1 has been associated with newly infiltrating monocytes in MS lesions [32]. CCR1-deficient mice show an attenuated EAE course [33]. Additionally, CXCR1, a pro-inflammatory molecule is upregulated by TCF2. Furthermore, TCF2 downregulated CD28, a co-stimulatory molecule involved in T-cell stimulation and IL6R, a pro-inflammatory molecule. Taken together, it is plausible that TCF2 plays a central role in the pathogenesis of MS.

Interestingly, a mutation in TCF2 has been linked to the etiology of MODY5 (Maturity-Onset of Diabetes, Type 5) and in a recent Danish nationwide cohort study, intra-individual and, to a lesser degree, intra-familial co-occurrence was evident in MS and type 1 diabetes. The underlying mechanisms may involve both genetic and environmental factors [34]. Individuals with type 1 diabetes are more than three times more likely to develop MS than controls; in addition, the two diseases appear to be linked, albeit to a weak extent, within families. The exact mechanism of how mutations in these transcription factors cause diabetes remains unknown. To reiterate, a link between TCF2 mutation and MODY5 has been established, while type 1 diabetics have an increased risk of MS development. Our study points to TCF2 having a prominent role in regulation of other transcription factors. In summary, the role of TCF2 in MS needs to be further explored since MODY5 and TCF2 are linked and type 1 diabetes confers increased risk of MS development.

Another transcription factor that we found in our transcription analysis, CDC5L, was downregulated. It is a well-characterized pre-mRNA splicing factor and involved in cell cycle kinetics in yeast, S. pombe; its role in MS, if any, is not characterized.