Robustness And Origin Of Cancer And Aging Related DNA Methylation Biomarkers

As an important mechanism of epigenetics, DNA methylation is closely associated with cancer development and progress. Recently, a large amount of cancer diagnostic DNA biomarkers have been identified based on the absolute DNA methylation levels. However, such biomarkers usually have low reproducibility. The rank-based cancer expression biomarkers have been proven to be robust for transcriptome analysis. Therefore, in this dissertation, we first evaluated the feasibility of finding cancer diagnostic biomarkers based on the relationship of ranks of the DNA methylation levels of CpG sites. We extracted reversal pairs of Cp G sites whose relative methylation level ranks significantly differ between cancer samples and normal controls from each DNA methylation profiling dataset respectively for five types of cancer. Then, based on the relative rank orderings of each biomarker pair, the majority vote rule was used to determine labels of samples in independent validation datasets. All biomarker pairs identified for each cancer type showed high classification accuracy(all above 90%) in independent datasets, indicating that the rank-based DNA methylation cancer diagnostic biomarkers were robust and can be easily applied to predicate individual samples.Comparing to biomarkers detected from cancer tissue, biomarkers detected from peripheral blood could be easily transferred to clinical application. However, the specificity and origin of such blood biomarkers are still unclear. We have analyzed the aberrant DNA mehylation CpG sites detected from head and neck carcinoma and ovarian cancer peripheral whole blood and head and neck carcinoma serum datasets, respectively. The results showed that more than 94.7% of the aberrant DNA methylation CpG sites respectively observed in each dataset significantly differed between myeloid lineage and lymphoid lineage cells and the alteration directions of DNA mehtylation levels of these CpG sites in cancer comparing with normal samples were consistent with that in myeloid lineage cells comparing with lymphoid lineage cells. The same phenomena were also observed in rheumatoid arthritis peripheral whole blood and inflammatory bowel disease peripheral leukocyte DNA methylation profiling datasets, respectively. These results indicated that the aberrant DNA methylation alterations observed in cancer and inflammatory disease patient peripheral whole blood and serum were determined by the population shifts of increased myeloid lineage cells and decreased lymphoid lineage cells, reflecting the differentially methylated signals between the two groups of cells. Therefore, DNA methylation biomarkers detected from cancer blood may be not specific enough to distinguish cancer from inflammatory diseases.Aging-related molecular alteration is one of the important risk factors for cancer. However, the origin of the aging-related DNA methylation alterations in peripheral whole blood is also unclear. Based on DNA methylation profiling datasets for peripheral whole blood, CD4+ T cells and CD14+ monocytes, we evaluated the origin of the aging-related DNA methylation sites(arCpGs) in peripheral whole blood. The results showed that the arCpGs observed in different peripheral whole blood datasets were significantly overlapped and consistent. The proportions of myeloid lineage cells and lymphoid lineage cells increased and decreased with age respectively but both weakly(Spearman rank correlation coefficients were only 0.15~0.22), thus only the CpGs with large DNA methylation level difference between myeloid lineage cells and lymphoid lineage cells could be identified as arCpGs in peripheral whole blood. The arCpGs observed in CD4+ T cells and CD14+ monocytes randomly overlapped(P=0.794), which also had different in- and out-CpG island distributions and were involved in different functional categories. The arCpGs observed in CD4+ T cells and in peripheral whole blood significantly overlapped with consistent positive or negative correlations with age(P<2.2×10-16). The arCpGs observed in CD14+ monocytes did not significantly overlap(P=0.365) with arCpGs observed in peripheral whole blood. However, 90.1% out of the 51 overlapped arCpGs showed consistent positive or negative correlations with age with those observed in peripheral whole blood and in CD4+ T cells, hinting that these arCpGs may mainly reflect the common alterations occurring in leukocyte subtypes. Thus, the arCpGs observed in peripheral whole blood mainly reflected the common or specific arCpGs of some leukocyte subtypes, but some reflected the differentially methylated CpG sites between myeloid lineage and lymphoid lineage cells due to the aging-related peripheral leukocyte proportion changes.In conclusion, we proposed a new method to detect the cancer diagnostic DNA methylation biomarkers from solid tissue and analyzed the origin of cancer and aging-related DNA methylation biomarkers detected from peripheral whole blood or serum, which have important implication in the identification of cancer biomarkers(especially those distinguishable from inflammatory diseases) and in the evaluation of cancer risk.