Genome - Epigenome Interactions Associated with Myalgic Encephalomyelitis/Chronic Fatigue Syndrome, 2017, McGowan et al

[Andy]

ABSTRACT

Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) is an example of a complex disease of unknown etiology. Multiple studies point to disruptions in immune functioning in ME/CFS patients as well as with specific genetic polymorphisms and alterations of the DNA methylome in lymphocytes.

However, the association between DNA methylation and genetic background in relation to the
ME/CFS is currently unknown. In this study we explored this association by characterizing the genomic (~4.3 million SNPs) and epigenomic (~480 thousand CpG loci) variability between populations of ME/CFS patients and healthy controls.

We found significant associations of methylation states in T-lymphocytes at several CpG loci and regions with ME/CFS phenotype. These methylation anomalies are in close proximity to
genes involved with immune function and cellular metabolism. Finally, we found significant correlations of genotypes with methylation phenotypes associated with ME/CFS.

The findings from this study highlight the role of epigenetic and genetic interactions in complex diseases, and suggest several genetic and epigenetic elements potentially involved in the mechanisms of disease in ME/CFS

Full text pre-print in PDF form - https://www.biorxiv.org/content/biorxiv/early/2017/12/22/237958.full.pdf

 

[Valentijn]

Not a convincing study, but perhaps a starting point. The number of subjects was very low (61 patients, 48 controls) for a genetic study, with millions of comparisons being made between SNPs and genetic expression. And then 10 were completely excluded for genetic reasons (too closely related, too different in ethnic background, technical issues with missing results, etc).

They also then did a (post-hoc?) re-analysis which excluded an additional 18 patients and 12 controls who had some overlap in health scores, for a total of 39 patients and 30 controls. Emotional well-being scores tend to be fairly normal for ME/CFS patients, so they may have turned a small difference into a much larger one by including it as a criteria for excluding patients if their scores were too normal. It could have the effect of primarily selecting for mood disorder in the re-analysis, in preference over ME/CFS symptoms or diagnosis:

In addition to these criteria, we re-analysed the data excluding individuals with health-related quality of life measurements that overlapped between cases and controls. This exclusion of intermediate illness phenotypes was aimed at increasing the power to detect possible associations between disease status and (epi)genotypes by decreasing the heterogeneity in phenotype symptomatology within each group. We utilized the scores of RAND-36 scales (physical functioning, energy/fatigue, emotional well-being, social functioning, pain, and general health) as quantitative measurements of ME/CFS phenotypes because these were significantly different (α = 0.05) between case and control groups, prior to excluding individuals with intermediate phenotypes.

Lymphocyte proportions were normal:

These observations suggest that alterations of general lymphocyte type proportions may not be a characteristic feature of ME/CFS. Rather, abnormalities in immune system functioning associated with ME/CFS appear to involve alterations in the activity and abundance of specific sub-populations.

In the original analysis, they did correct for making millions of comparisons and did not find any associations. But the study probably didn't have enough participants to find an association anyhow:

None of the more than 2 million variable SNP loci targeted in this study had a significant association (α= 0.05) with ME/CFS after p-value corrections with Bonferroni, Holm, Benjamini and Hochberg, or permutation methods when data from both sexes were analysed together.

They then repeated the analysis with only female patients and controls, and rs11712777 came up as barely being significant. This is likely a fluke, however, since even the minor allele is pretty common (up to 39% MAF in Europeans), and the SNP is not on a gene, and quite far from any genes (40,000-60,000 SNPs). One of the closer genes is CCK, a receptor for which is associated with sickness behavior, and there are SNPs which rs11712777 is in strong linkage disequilibrium with (people often inherit the two SNPs together) and can impact gene function somewhat. But there's no explanation for the apparent lack of significant difference between patients and controls on the linked SNPs.

It sounds like there weren't any significant epigenetic results, after corrections for multiple comparisons were applied, though there were 100+ prior to corrections being made. It is a bit interesting that some of the uncorrected results were in line with previous results (assuming different patients were used):

We found 31 genes associated with DMPs in T-cells that were common to this study and a previous study by our group. These genes, which include PAX8 (paired box 8), and ATP4B (ATPase H+/K+ transporting beta subunit) (Supplementary Table S1), are involved in the regulation of cellular processes and cell signaling. This is in line with recent ME/CFS work that observed differences in cellular metabolism in ME/CFS.

So it looks like this is a null result under the proper analysis, but it was too underpowered to find much that way. The uncorrected secondary analyses found some differences, but many or all of those results are likely to be false positives due to comparing so many variables. I don't think anything can be concluded from this study, but it could serve as a starting point for more focused analysis in small samples. Given the lack of power resulting from the low number of participants, I wish they'd used this study to verify their earlier work by focusing on a very small number of earlier identified SNPs and/or gene products, rather than going on another fishing expedition which also now needs replication.