In this study of adults with variable asthma severity, we identified extensive genome-wide differences in DNA methylation in CD3 + T cells of individuals with asthma that correlated with markers of airway inflammation, lung function, and asthma control. The differences in DNA methylation were widespread throughout the genome but varied depending on the comparison. Many of the genes found to be differentially methylated are known to be involved in antigen presentation or modulation of T cell differentiation, supporting the potential importance of DNA methylation changes in asthma. Although some genes have been shown in the literature to be clearly associated with asthma, other genes are novel, potentially providing new insights into asthma pathophysiology.
Changes in DNA methylation have been well-described in children who develop asthma [1, 2, 5, 6] and represent a fundamental mechanism that support the fetal-origin hypothesis of the disease. However, asthma is a complex disease where patients develop asthma at different ages of onset and exhibit varying levels of airway inflammation and severity. Although changes in DNA methylation are well-associated with asthma development [1–3, 6, 10], the role of DNA methylation in modulating asthma severity is less appreciated. By examining multiple facets of asthma severity and phenotype including FeNO, pre-bronchodilator FEV1, and symptom control by ACT, we were able to identify methylomic changes that were associated with each of these clinical variables. That methylation differences with each of these comparisons produced a unique set of differentially methylated probes and genes speaks to how each of these variables – FeNO, FEV1, and ACT – serve as different measures of asthma severity and pathophysiology. That many of the identified differentially methylated genes were different from those described in studies of children highlighting potential differences between adult and childhood asthma pathophysiology and the role methylation plays in asthma development vs severity.
As compared to many other DNA methylomic studies in asthma that utilize whole blood or peripheral blood mononuclear cells, we specifically isolated CD3 + T cells for our study, as DNA methylation patterns are often cell-type specific and identifying changes in methylation attributable to a specific cell type may provide greater insights into the significance of these changes. Because anti-CD3 + magnetic bead isolation is subject to variability in technique, we additionally analyzed samples by cell-type deconvolution, which showed that although our samples consisted mostly of CD4 and CD8 T cells, a small percentage included NK cells, B cells, monocytes, and neutrophils. Our model thus used cell-type deconvolution as a covariate. Although other reference-free methods for cell-type deconvolution [35] could also have been used, the combination of CD3 + magnetic bead isolation and the cell-type deconvolution we employed was able to identify methylation differences with genes involved in antigen presentation and T cell function and activation. More specifically, methylation differences between those with high and low FeNO, which is often used clinically as a marker of type 2 airway inflammation, were associated with methylation differences in IL4, CCL11 (eotaxin), and many HLA genes that are involved in antigen presentation and MHC class II assembly. Coincidentally, the cutoffs between subjects with high vs low FeNO in our cohort fell near 25 ppb, which is the cutoff commonly used in clinical guidelines [36] to distinguish type 2 vs non-type 2 asthma, suggesting that these methylation differences may be important in distinguishing different clinical phenotypes or endotypes of disease. The lack of available data on subjects’ peripheral eosinophil levels, total IgE level, or specific aeroallergen sensitivity testing is however, a limitation and prevents extrapolation of our findings from FeNO with a true “type 2 high” phenotype.
Although comparisons of methylation patterns between subjects with high vs low FEV1 also resulted in methylation differences in genes associated with T cell function and activation, there were also other genes that were differentially methylated, including many related to drug metabolism and cytochrome P450. The significance of this is unknown but may indicate a mechanism by which environmental toxins can affect asthma and influence lung function. Other genes identified as differentially methylated in the high vs low FEV1 comparison include those involved in inflammation and are downstream targets of JAK2 and TNF. These include genes such as IL6, interferon γ (IFNG), SOCS3, and NOS3. Complement genes (C3 and C4A) and genes involved in TH17 differentiation were also highlighted. These findings indicate how much methylation differences in these genes that regulate and modulate inflammation may influence lung function. Finally, it was interesting to note that individuals with asthma with different levels of symptom control, as measured by ACT, had differential methylation in yet other genes, some of which were involved in development and airway remodeling, such as HOX family genes, integrins (ITGB5, ITGA6), desmin (DES), and endothelin (EDN1), suggesting that differential methylation of these genes may have an important role in daily symptoms and asthma control. As asthma patients with diminished lung function and persistent symptoms often exhibit pathologic evidence of airway remodeling, the differential methylation of these inflammatory and developmental genes may predispose patients to developing airway remodeling.
Some of the genes that were differentially methylated were observed to be differentially methylated in other studies of asthma as well, such as RUNX3 and IL4 [6]. This supports the generalizability of our findings. Certain genes, however, were found unique to our study but differentially methylated across all our comparisons. These include certain HLA family genes, PTPRC, and PTPRJ. These latter two genes are protein tyrosine phosphatases that have been described as master regulators that are differentially expressed in transcriptomic studies of children with severe asthma [34, 37]. These genes have also been shown by others to be regulated by DNA methylation and our study is the first to identify the differential methylation of these genes to be associated with asthma severity. AGT, which codes for angiotensinogen, was another gene present in the center of many of our network analyses and shown among all our clinical comparisons as differentially methylated.
Our study was unique in that we studied a cohort of asthma patients from China, as compared to other population-based DNA methylomic studies that focus on European or North American cohorts. Differences in local environment and ancestral origin may account for some of the differences in DNA methylation in our study compared to others though there were many genes and pathways that were also similar. We excluded patients with smoking history to eliminate potential confounding effects of patients with COPD or asthma-COPD overlap. Although the total size of our cohort was small, longitudinal collection of blood samples from the same patient enhanced the power of our study and all analyses were adjusted for repeat measurements from the same patient.
One of the limitations of the study was that we did not observe robust longitudinal changes in DNA methylation that correlated with a given subject’s FEV1, FeNO, or ACT over time, as we initially set out to do. Instead, the DNA methylation patterns for each patient remained relatively stable, and greater variations in DNA methylation were present between patients as compared to changes in methylation over time within subjects. This suggests that DNA methylation may be a stable epigenetic mark though this may also be a consequence of the fact that the degree of asthma severity in most of the subjects remained relatively stable throughout longitudinal follow-up and no subjects developed acute exacerbations during the study period. Not surprisingly then, attempts to linearly correlate DNA methylation patterns with absolute values of FeNO, FEV1, or ACT was poor. This, however, did not preclude our ability to find major differences in DNA methylation when dichotomous groups of high vs low FeNO, FEV1, or ACT were compared. These findings suggest that DNA methylation may be responsible for fundamental differences in baseline asthma phenotype/endotype but it is unclear if it is predictive of day-to-day asthma variation. Future studies on a larger cohort or that focuses on subjects with more variable asthma control will allow us to examine whether DNA methylation might also change over time in patients with labile disease.
Although we observed robust patterns of differential methylation, the biological significance of these differences in methylation is unknown. In our comparisons, we observed shifts in the distribution of methylation differences, with more hypomethylation in CpG islands and shores, for example, among subjects with high FeNO. This suggests that there may be an increase in gene expression among patient with high FeNO, which would be congruent with the presence of greater airway inflammation, though this remains to be determined experimentally. To summarize the net effect of methylation at the gene level, we calculated the weighted average of the methylation changes of all CpG sites for a given gene while maintaining the minimum FDR. Calculating the average methylation for a gene has the advantage of avoiding bias that may occur with genes that possess variable number of CpG sites assayed, uneven distribution of CpG sites, and inherent genomic biases within the Epic array. However, the effects of methylation are complex and often context dependent and this approach of taking the weighted average may obscure important methylation differences that only occur in specific genomic regions. Furthermore, this approach does not completely remove bias from genes that possess a greater number of CpG sites and sequencing-based approaches such as whole genome bisulfite sequencing, though more costly, would provide a more comprehensive picture of methylation differences. Gene pathway enrichment analysis was performed over the background of all genes analyzed by the Epic array, but ultimately, the functional consequences of the observed methylation changes are unknown. The genes, networks, and pathways that we highlight in this study require future studies to determine their mechanistic relevance.
In conclusion, differential patterns of DNA methylation were observed in the peripheral blood T cells of patients with different degrees of asthma severity. These differences occurred in many genes associated with T cell function, antigen presentation, and leukocyte activation, indicating a potential role for DNA methylation to modulate the immunologic pathways critical to asthma. Novel differentially methylated genes were also identified that may influence T cell function or affect asthma pathobiology and warrant future studies. Together, these data support the potential for this epigenetic mechanism as a critical determinant of asthma phenotype and severity.