Influence of genetic background and tissue types on global DNA methylation patterns

Abstract

Recent studies have shown a genetic influence on gene expression variation, chromatin, and DNA methylation. However, the effects of genetic background and tissue types on DNA methylation at the genome-wide level have not been characterized extensively. To study the effect of genetic background and tissue types on global DNA methylation, we performed DNA methylation analysis using the Affymetrix 500K SNP array on tumor, adjacent normal tissue, and blood DNA from 30 patients with esophageal squamous cell carcinoma (ESCC). The use of multiple tissues from 30 individuals allowed us to evaluate variation of DNA methylation states across tissues and individuals. Our results demonstrate that blood and esophageal tissues shared similar DNA methylation patterns within the same individual, suggesting an influence of genetic background on DNA methylation. Furthermore, we showed that tissue types are important contributors of DNA methylation states.

Figure 1. Analysis of DNA methylation patterns in blood and esophageal tissue from 30 ESCC patients.

Figure 1A shows the PCA using methylation measurements. Two clusters are evident, corresponding to blood and normal samples. Samples are labeled with blue (patient blood, letter B) and green (patient normal, letter N). The numbers indicate patients. Blood and normal samples from a single patient are connected by a dashed line. The dashed lines are mostly parallel to PC2 axis due to nearly identical PC1 scores between blood and normal esophageal tissue from a single individual. Figure 1B shows unit length direction arrows of the same data as Figure 1A. An arrow emanates from a normal sample and points to blood. The direction of the arrow is identical to that of the dashed line in Figure 1A. Figure 1C shows similar PCA using quantitative DNA measurements. Data processing and analysis is similar to Figure 1A. The DNA measurements serve as a control for variation in the samples. Figure 1D shows the plot of arrows using the data from Figure 1C. It is evident that the arrows in Figure 1D (for DNA) are randomly orientated due to different PC1 scores between the two tissues from the same single individual.

Figure 2. Analysis of DNA methylation patterns in blood, normal esophageal tissue, and tumors.

The analysis is similar to Figure 1 except that we include 30 tumors from the same individuals. Figure 2A shows the PCA using methylation measurements. Three clusters are evident, corresponding to blood, normal, and tumor samples. Samples are labeled with blue (patient blood, letter B), green (patient normal, letter N), and red (patient tumor, letter T). Figure 2B shows unit length direction arrows of the same data as Figure 2A. The green arrows emanate from normal and point to blood samples whereas the red arrows start from tumors and point to blood samples. Figure 2C shows similar PCA using quantitative DNA measurements. Data processing and analysis is similar to Figure 2A. The DNA measurements serve as a control for variation in the samples. Figure 2D shows the plot of arrows using the data from Figure 2C. It is evident that the arrows in Figure 2D (for DNA) are randomly orientated due to different PC1 and PC2 scores between different tissues from a single individual.

Figure 3. Heat map displays pair wise correlation results for methylation and genomic DNA data.

Figure 3A contains methylation data. We have 90 microarray data generated from samples consisting of 3 tissues (blood, normal, and ESCC) from 30 individuals. All pair wise comparisons were analyzed, and Pearson correlation coefficients were plotted in the heat map. Figure 3B has genomic DNA data. Analyses are similar to Figure 3A except for the use of quantitative values from the genotype experiments.

Figure 4. Analyses of DNA methylation difference between blood and normal esophageal tissue for single individuals.

Figure 4A displays the genes that exhibit similar methylation measurements between blood and normal esophageal tissue for single individuals relative to the differences between blood and normal tissue for different individuals. Each circle represents a comparison of: (1) methylation measurements between blood and normal tissue (blue from the same individual labeled as Hpa2.paired, 30 data points per gene; red from two different individuals labeled as Hpa2.unpaired, 870 data points) or (2) DNA analysis (green from the same individual labeled as gDNA.paired, 30 data points; pink from two different individuals labeled as gDNA.unpaired, 870 data points). These calculations were carried out for the genes indicated on the x-axis. The SNPs marking the individual genes are rs7203335 (SNX29), rs8190404 (DIA1), rs12780199 (INPP5A), rs3760220 (DKFZP586L0724), rs3768723 (PIGF), rs7605146 (DNAPTP6), rs11254 (ETS2), rs6464151 (PRKAG2), rs2302592 (TTLL5), and rs1265074 (CCHCR1). Figure 4B shows the histogram of p-values in the Ansari-Bradley two-sample test for methylation (red bars, labeled with Hpa2) and DNA (green bars, labeled with gDNA) measurements. For each SNP, the ratio of scales was tested for two samples: one is the methylation differences between blood and normal tissue from 30 pairs of individuals and another is the methylation differences between the two tissues from two different individuals from the 870 pairs resulting from selecting 2 out 30 individuals in all combinations. As a control, we performed similar analyses using data from DNA array experiments. The histogram summarizes the distribution of the negative log10pvalues from the analyses of the methylation and DNA data.