We present the first report of genome-scale analysis of DNA methylation profile in JIA. We have identified 145 differentially methylated loci of which 91 are more highly methylated in cases. Significant case-control differences are generally modest (difference in median β ranging from 0.006 – 0.19), however, this is consistent with a growing number of other equivalent studies in complex disease (reviewed in
). Pathway analyses of genes subject to altered DNA methylation identified autoimmune disease-relevant networks of genes including ‘immunological disease’, ‘cellular growth and proliferation’, and ‘antigen presentation’.
Four case samples used in the HM27 analysis had prior exposure to the commonly used DMARD, MTX. The anti-inflammatory properties of MTX are yet to be fully understood, but may include inhibition of T cell activation, induction of T cell apoptosis, and/or alteration of expression of cytokines
. Importantly for DNA methylation, MTX is an anti-folate agent. Folates are micronutrients essential for one-carbon metabolism, a process of methyl group transfer essential for the specific methylation of cytosine nucleotides within CpG sites
. Thus, it is reasonable to hypothesise that MTX may impact the T cell methylome in exposed cases. Whilst we cannot, from our data, provide conclusive evidence in support of this hypothesis, removal of MTX-exposed cases from the array analysis did have a striking impact on the outcome: the number of differentially methylated genes was reduced from 145 to 11; and none of the genes from the first- and second-tier analyses overlapped. While it remains possible that the contrasting outcomes were due to other factors, for example, the MTX-exposed cases may have been in other ways clinically distinct from the MTX-naive cases, our data does suggest a potential impact of MTX on DNA methylation. However, we found no evidence to suggest that this impact is genome-wide since mean β values across the >25,000 HM27 probes analysed was not different between exposed and unexposed cases (t-test p = 0.85). Unfortunately our sample size was insufficient for a meaningful locus-by-locus comparison of MTX-exposed and -naive case samples within this study, but overall our work suggests that data derived from MTX-exposed samples should be treated with caution in epigenome-wide association studies. Further investigation of this issue using a larger sample of MTX naive and exposed individuals is warranted, especially in light of recently published rheumatoid arthritis methylome data derived from case samples of mixed MTX exposure status
Pathway analysis based on the 11 MTX-naive differentially methylated genes defined a network related to ‘cellular growth and proliferation, haematological system development and function, hematopoiesis’ within which nine of the 10 characterised genes were placed. Central to this network is the pro-inflammatory cytokine TNF. TNF initiates a pro-inflammatory programme of events associated with autoimmune and inflammatory diseases
, including JIA
. TNF blockers have been used for some time as a treatment for JIA
. Therefore, aberrant methylation of a network of differentially methylated genes in T cells of children with JIA may result in the promotion of a disease-relevant inflammatory cascade.
Of the MTX-naive differentially methylated loci, IL32 stands out as an interesting finding. IL32 is a pro-inflammatory cytokine shown to be a potent inducer of TNF. TNF also induces IL32 expression and the two molecules form an important autoinflammatory loop, as demonstrated in RA synovial cells
. Mice overexpressing IL32 show higher serum concentrations of TNF, and exacerbation of collagen-induced arthritis
. In our study, specific IL32 CpGs measured by both genome-scale and locus specific analyses were seen to be less methylated in cases than in controls across two distinct sample sets, suggesting higher levels of expression. A limitation of our study is that we did not collect CD4+ T cell RNA to measure gene expression, although evidence does exist to suggest IL32 expression is dependent on methylation
. Thus, further work will be required to assess the relationship between IL32 DNA methylation and IL32 gene expression in JIA.
The locus-specific replication data for MRPL28 is also interesting. Although MRPL28 was not seen to be significantly differentially methylated in the MTX-naive array analysis (FDR p = 0.47), significantly reduced methylation in cases relative to matched controls, (Δβ values in excess of 0.2), was found for pairs containing both MTX-exposed and -naive cases across discovery and replication samples. MRPL28 encodes mitochondrial ribosomal protein L28, the function of which is poorly described. It is therefore difficult to speculate as to how reduced MRPL28 methylation might increase JIA risk, although mitochondrial dysfunction has been associated with autoimmune disease
. Again, further work will be required to more fully interpret these findings.
An important issue that needs to be addressed in epigenetic research such as that presented here, is whether or not any identified disease-associated changes to the epigenome precede disease onset, or result from the disease process. Identification of aberrant DNA methylation that precedes disease suggests that such changes may contribute to risk of developing disease, and could therefore act as biomarkers to identify at-risk individuals and contribute to our understanding of the genomic underpinnings of causality. However, for relatively rare complex diseases like JIA, it is very difficult to prospectively collect cases, since the size of the prospective cohort necessary to identify sufficient cases would be very large. One approach to the problem is to retrospectively obtain diagnostic biospecimens collected routinely at birth, such as neonatal dried blood spots. Methods have now been developed to measure locus-specific DNA methylation in dried blood spots
, and in the future this may well extend to the use of these specimens on a genome-scale. The CLARITY Biobank now encompasses neonatal dried blood spots from recruited cases and controls, and is shortly commencing longitudinal biospecimen collection for cases across the disease course. Thus, a future important research direction will be to examine changing methylation levels periodically from birth through to disease resolution.