We have compared gene expression in relapse and remission in paired samples from 14 MS patients using microarrays which cover the known expressed mRNA in the human genome. We found significant changes in expression of several mRNA, which may clarify the nature of the disease process in MS. Several features of these results deserve comment.
First, the most salient finding is that there were changes in both T cells and non-T cells. The changes in non-T cells were primarily decreases in selected transcripts. Not all non-T transcripts were decreased, so this does not appear to be due to migration of these cells out of the blood in relapse, but rather to a change in the expressed genes. Where their function is known, many of the decreased transcripts have a regulatory or inhibitory role. CDKN1C
are both likely inhibitors of cell division, IL1RN
inhibits the activity of interleukin-1, and PILRA
has a tyrosine-based inhibitory motif and regulates the protein tyrosine phosphatase SHP-1, which is central in the control of cell signaling and may be altered in MS (12
). The downregulation of mRNA for inhibitory proteins suggests that the non-T cells are becoming activated. Further study is needed to define the types of non-T cell involved.
In contrast, mRNA differentially expressed in T cells were increased predominantly in expression. The ones with the most significant changes in expression are involved in metabolic processes or are of unknown function. The most significant change in a cytokine mRNA is the increase in IL16, which is known to be chemotactic for cells bearing CD4. Other cytokine mRNA changes are decreased IFNG, IL6, and IL15.
Second, the changes in expression are modest in size, in the range of 70% to 130% of the baseline values. There are several potential explanations for this. It is probable that only a fraction of circulating PBMC are involved in the immunologic events causing a clinical relapse. Also, the disease process in MS is more continuously active than the clinical symptoms indicate, so many of our stable samples probably have similar changes. Finally, we used PBMC, which contain a variety of different cell types, so changes in mRNA expression in a single cell type may be obscured by expression of the same mRNA in other cell types.
Third, the changes in mRNA expression did not necessarily predict changes in protein concentration in serum. Multiple factors could contribute to this, including post-transcriptional regulation, rates of protein secretion into serum, the half-life of the protein in serum, and cell types other than PBMC secreting particular proteins. The microarray results are better interpreted as an indicator of the internal state of the PBMC rather than the external milieu in the serum.
There have been several previous studies of expression changes in relapse, and we compared transcripts reported to be altered in those studies with our set of 530 transcripts altered at a significance level of 0.01. The concordance between our results and previous studies is limited. This is not unexpected given differences in methods. The majority of previous studies compared small numbers of patients in relapse to a different group of stable patients. The large variability of gene expression between individuals seen in this study () suggests that any changes due to relapse are likely to be obscured. Only two studies used paired samples, and both of those studies tested only T cells positively selected by anti-CD3 antibody. This approach discards the information from non-T cells, which proved to be interesting in our study. It also might cause changes in mRNA expression in the T cells due to crosslinking of CD3, with unknown effects on the results. In this work, we isolated the PBMC and stabilized RNA as quickly as possible to minimize ex vivo changes in mRNA expression.
The most directly comparable study is that of Satoh et al
). This study included only six subjects with paired samples, used a relatively small microarray with 1,258 selected transcripts, and positively selected T cells with anti-CD3 antibody. They identified 43 differentially expressed genes, none of which were altered in this study. Achiron et al
. have reported two studies, which compared unpaired samples, some from patients in relapse and others from different patients in remission (5
). We replicated their findings that BCL2
is decreased and BNIP3
is increased. More recent work from the same group reports that expression of a small set of mRNA can be used to predict relapse (14
), but the transcripts they used as predictors were not altered in our patients. Arthur et al
. also reported a comparison of unpaired relapse and remission patients (6
), but comparison with their results is difficult since they compared relapse to controls and remission to controls rather than directly comparing relapse and remission. They found decreased CDKN1C
in both relapse and remission, and an increase in OGT
in relapse, but not remission. Malmestrom et al
. compared expression in 10 subjects with paired relapse and remission samples (7
), but they used anti-CD3 antibody to positively select T cells, and their reported results cover only a limited range of transcripts and proteins in CD8 cells with no overlap with our findings. Finally, we found no overlap with the findings of Brynedal et al
., who compared expression from unpaired blood samples from 10 stable patients and 14 patients in relapse (15
). The poor agreement of our study and previous work is likely due to differences in microarrays, differences in the cell types used as the source of mRNA, the use of paired rather than unpaired samples, the limited results reported from some studies, and small sample sizes leading to false positive results.
The concordance of our pathway analysis with previous studies is likewise limited. We found that the NF-κB pathway was changed in agreement with Satoh et al
). We found changes in some transcripts involved in apoptosis as discussed above, but we did not find significant alterations in apoptotic pathways as reported by Achiron et al
. and Arthur et al
We found better concordance with studies comparing MS patients and controls (). It is reasonable to expect that some of the mRNA changes that distinguish MS patients from controls might also distinguish MS relapses from remissions. Indeed, we found concordant changes in many transcripts which had previously been reported to be changed in MS (16
). Of particular interest is the large number of concordant findings with the study of Corvol et al
). This is somewhat unexpected, since they studied gene expression in negatively selected CD4 T cells in patients with clinically isolated syndrome. This suggests that transcripts altered early in disease also change in relapse. Since their CD4 cells were 95% pure, and they still measured changes in transcripts expressed primarily in non-T cells, this raises the question of whether they are finding changes in the residual non-T cells or if these transcripts also are altered in CD4 cells. In addition to gene expression studies, we compared our results to the findings of genome-wide association studies in MS (20
). Three of the loci identified in recent studies, IL7R
, had changes in expression during relapse.
The three proteins we studied all have potential relevance to MS. C1Q has immune regulatory properties and decreases in C1Q are associated with autoimmunity (11
). Although we could not demonstrate a change in serum C1Q in our 19 patients, others have measured a decrease in C1Q and other complement components in a larger group of MS patients in relapse (22
). IL-16 has been implicated in autoimmune diseases (23
), including MS and animal models of MS (24
), and IL-16 is downregulated during interferon treatment of MS (26
). IL-1RA has antiinflammatory activity, and decreases in IL-1RA would be expected to cause inflammation. The role of IL-1RA genotype as a risk factor for MS is controversial (27
), but IL-1RA protein is increased during interferon or glatiramer treatment (28
). The only previous study measuring IL-1RA protein in relapse found it was increased rather than decreased (30
), but they compared IL-1RA levels in eight relapsing patients to a different group of stable patients.
One consideration in interpreting these results is that we tested patients only at two time points—the onset of clinical symptoms and while clinically stable. The hypothesized immune activation in the periphery should occur before inflammation in the CNS. By the time symptoms appear, the relevant changes in the peripheral blood may have declined or even resolved. One prospective study with blood samples done every 2 to 4 weeks found that stimulated IFN-γ production in vitro
peaked 1 to 2 weeks before clinical symptoms, and continued to decline after the onset of symptoms (31
). This is consistent with our observation of decreased IFNG
mRNA at the time of onset of symptoms. Some of the changes measured in blood at the onset of symptoms may reflect the resolution of recent immune activation.
In conclusion, the observed changes in gene expression in PBMC with MS relapse suggest an activation of non-T cells and changes in function in T cells. Further investigation of the events occurring with MS relapse should be pursued. The optimum study would include more subjects, perform serial sampling of patients at frequent time intervals, use MRI rather than clinical relapse as the measure of MS disease activity, include a white cell differential to document the proportion of monocytes and lymphocytes and study more homogeneous populations of leukocytes.