1,958
Views
15
CrossRef citations to date
0
Altmetric
Research Paper

Epigenome-wide cross-tissue correlation of human bone and blood DNA methylation – can blood be used as a surrogate for bone?

, , ORCID Icon &
Pages 92-105 | Received 17 Feb 2020, Accepted 28 May 2020, Published online: 21 Jul 2020

Figures & data

Figure 1. Schematic illustration of the experimental design and bioinformatic analysis strategy to determine the correspondence between bone and blood methylomes

Figure 1. Schematic illustration of the experimental design and bioinformatic analysis strategy to determine the correspondence between bone and blood methylomes

Figure 2. Beta-value density plots of the raw and preprocessed data for bone and blood

Plots are coloured by batch (Red, Batch-1; Blue, Batch-2). Batch effect is observed in the raw data, but removed by the preprocessing methods. In the bottom plot, bone and blood show distinct methylation profiles, with bone having a broader distribution around higher methylation levels.
Figure 2. Beta-value density plots of the raw and preprocessed data for bone and blood

Figure 3. Multi-dimensional scaling plots of the M-values in the raw and preprocessed data for bone and blood

Data points are BN (bone) and BL (blood) with patient ID, coloured by batch (Red-Batch 1; Blue-Batch 2). Tissue type constitutes the main variance in the data; bone and blood samples are distinctly grouped in the raw and preprocessed data. Within the tissues, batch effect is evident in the raw and normalized data based on clustering of the samples, but is removed after batch correction.
Figure 3. Multi-dimensional scaling plots of the M-values in the raw and preprocessed data for bone and blood

Figure 4. Enrichment of DMPs according to genomic and CpG island coordinates

The bars show the distribution of differentially methylated CpG sites and the ratios of hypo- and hyper-methylated sites in various regions. The number of DMPs identified and total number of CpG sites in the preprocessed data are also reported. DMPs are depleted in promoter regions and CpG islands. The sum of DMPs in genomic coordinates (n = 16,228) exceeds the total number of DMPs identified (14,625) since some sites have multiple annotations that refer to gene isoforms.
Figure 4. Enrichment of DMPs according to genomic and CpG island coordinates

Figure 5. Permutation analysis on the correlation testing to identify similarly methylated positions

Figure 5. Permutation analysis on the correlation testing to identify similarly methylated positions

Figure 6. Enrichment of SMP according to genomic and CpG island coordinates

The bars represent ratios of the CpG sites similarly methylated in bone and blood. For each region, the number of the SMPs identified and a total number of CpG sites in the preprocessed data are reported. SMPs are enriched in CpG islands and regions that have regulatory roles in gene expression. The sum of SMPs in genomic coordinates (n = 34,785) exceeds the total number of SMPs identified (28,549) since some sites have multiple annotations that refer to gene isoforms.
Figure 6. Enrichment of SMP according to genomic and CpG island coordinates

Table 1. Number of loci for bone phenotypes extracted from the GWAS catalogue (A), number of GWAS loci represented among the similarly methylated positions

Figure 7. Methylation levels in bone versus blood samples in the SMPs overlapping with 3 examples of genes involved in pathways known to be critical to bone biology – ESR1, EN1, and Wnt16.

The legends show the sites names and their genomic and CpG island coordinates. Some of the sites show high inter-individual variation, with methylation levels for bone and blood varying considerably between subjects, whereas other sites have low inter-individual variation and methylation levels are fairly consistent between individuals.
Figure 7. Methylation levels in bone versus blood samples in the SMPs overlapping with 3 examples of genes involved in pathways known to be critical to bone biology – ESR1, EN1, and Wnt16.

Table 2. Pathway enrichment analysis using the molecular signatures database (MSigDB) on all the SMPs (FDR q-value of correlation< 0.05, and Δβ<0.2)

Table 3. Pathway enrichment analysis using the molecular signatures database (MSigDB) on a ‘core’ set of highly significantly correlated SMPs (FDR q-value of correlation< 0.005, and Δβ<0.2)

Supplemental material

Supplemental Material

Download Zip (2.3 MB)