Integrative analysis identifies DNMTs against immune-infiltrating neutrophils and dendritic cells in colorectal cancer

Abstract

Molecular characterizations, including microsatellite instability (MSI) and the CpG island methylator phenotype (CIMP) showed strong associations in colorectal carcinoma (CRC) and provided a deeper understanding of the etiology of disease. However, the global relationship between epigenetic alternations and changes in mRNA expression in CRC remains largely undefined, especially regarding the roles of DNA methyltransferases (DNMTs). Here, we conducted a systematic network comparison to explore the global conservation between co-expressed and co-methylated modules. We successfully identified immune-related modules that were regulated by DNMTs and had strong associations with immune-infiltrating neutrophils and dendritic cells in CRC. Moreover, we found that genes in those modules were prognostic for CRC, with 97.1% (168/173) being significantly influenced by DNMTs. Thus, this study resolved an interaction between DNA methylation and mRNA expression through DNMTs. Additionally, we provided evidence that DNMTs control the global hypomethylation of oncogenes, including ALOX5AP and CSF3R that otherwise have high methylation in normal colons. Such genes were also more sensitive to DNMT changes, such as in CRC. Collectively, our analyzes provided a systems biology approach to investigate the association among different molecular phenotypes in diseases.

Keywords: Colorectal carcinoma; DNMTs; WGCNA; dendritic cell; neutrophil.

Figure 1.

WGCNA resolves co-expressed modules in CRC. (a) Cluster dendrogram of mRNA expression in 7,716 genes across 214 tumor samples established on topological overlap (TO) distance. The y-axis height corresponds to the distance (1-TO). The x-panel describes modules defined by dynamic tree cutting and the grey color represents genes not belonging to any modules. (b) Heatmap of modules correlated with molecular features of CRC. Modules are outlined based on their correlation with molecular traits (exp, expression; mut, mutation; cna, copy number alteration) when adjusted for a P-value < 0.05.

Figure 2.

WGCNA reveals co-methylated modules in CRC. (a) WGCNA clustered 7,716 genes by promoter DNA methylation values across 214 tumor samples based on topological overlap (TO) distance. The y-axis height corresponds to the distance (1-TO). The x-panel describes modules defined by dynamic tree cutting and the grey color indicates genes not belonging to any co-methylated modules. (b) Heatmap of modules correlated with molecular features of CRC. Modules are outlined based on their correlation to molecular phenotypes (exp, expression; mut, mutation; cna, copy number alteration) when adjusted for a P-value < 0.05.

Figure 3.

Module preservation in expression and methylation modules. (a) Bar graphs of the Z-summary statistic in co-expression networks using the methylation data set as the reference. The interpretation threshold for strong evidence (Z-summary > 10), moderate to weak evidence (2 < Z-summary < 10), and no evidence (Z-summary <2). (b) Bar graphs of the Z-summary statistic in co-methylation networks using the expression data set as the reference. The interpretation threshold for strong evidence (Z-summary > 10), moderate to weak evidence (2 < Z-summary < 10), and no evidence (Z-summary <2). (c) Co-expression patterns of genes overlapping in selected immune related modules. The red lines indicate the positive correlation between the genes (black dots), while line thickness indicates the strength of co-expression. The size of the dots represent the total connectivity of each gene. (d) Co-methylation patterns of genes overlapping in selected immune related modules.

Figure 4.

DNMTs influence intra-tumoral immune cells in colorectal cancer. (a) Dot plot of functional comparisons between conservative modules. The gene ratio represents the percentage of genes enriched in each significant GO-term. (b) Heatmap of each infiltrate correlated with immune-related molecular features of CRC. Modules are outlined based on their correlation with immune traits (exp, expression; mut, mutation; cna, copy number alteration) when adjusted for a P-value < 0.05. (c) Kaplan-Meier survival plot of two tumour groups classified by the abundance of infiltrating neutrophil. The P-value was calculated using the log-rank test. (d) Kaplan-Meier survival plot of two tumour groups categorized by the abundance of infiltrating dendritic cells. (e) Kaplan-Meier survival plot of two tumour classes grouped by the mRNA expression of DNMT1.

Figure 5.

Comprehensive analysis of survival-associated genes in the neutrophil and dendritic cell sets. (a) Of the 306 genes in the expression modules of neutrophils and dendritic cell sets, 173 were significant in survival analyzes. (b) Upset plot of the interactions among six groups of the DNMT system (DNMTs cna mRNA, mRNA expression of genes significantly affected by the copy-number alterations of DNMTs; DNMTs cna methy, promoter DNA methylation of genes significantly affected by copy-number alterations of DNMTs; DNMTs exp mRNA, mRNA expression of genes significantly correlated to mRNA expression of DNMTs; DNMTs exp methy, promoter DNA methylation of genes significantly correlated to mRNA expression of DNMTs; DNMTs mut mRNA, mRNA expression of genes significantly affected by somatic mutations in DNMTs; DNMTs mut methy, promoter DNA methylation of genes significantly affected by somatic mutations in DNMTs). Boxplots of DNA methylation among significant, NS, and other module groups in normal adjacent tissue (c) and CRC (d). The star indicates statistical differences based on the Wilcoxon rank-sum test. (e). Boxplots of the mRNA expression comparison between significant and NS groups in lymph node and colon.

Figure 6.

Network properties of survival-associated genes in neutrophil and dendritic cell sets. (a) Boxplots of module membership for comparison of survival-significant and NS genes in neutrophil and dendritic cell sets. The asterisks indicate the statistical difference based on the Wilcoxon rank-sum test. (b) Cytoscape network display for genes in the neutrophil and dendritic cell sets. The size of the nodes indicates the module membership of each node and the color corresponds to the gene group (red, significant and grey, NS).

Figure 7.

ALOX5AP and CSF3R were associated with oncogenesis and prognosis of CRC. (a) ALOX5AP had decreased promoter DNA methylation (N = 125 tumors and N = 29 normal tissues;unpaired t-test, P < 0.05, *; P < 0.01, **; P < 0.001, ***) and elevated expression (N = 26 tumors and N = 26 normal tissues;unpaired t-test) in CRC. When compared in DNMTs groups, ALOX5AP showed significantly downregulated methylation and upregulated expression in samples with DNMT mutation (N = 28 mutations and N = 186 non-mutations;unpaired t-test). (b) CSF3R showed decreased promoter DNA methylation and elevated expression in CRC. When compared in DNMTs groups, CSF3R had significantly downregulated methylation patterns and upregulated expressions in samples with DNMTs mutation. (c) Pearson correlations of ALOX5AP expression with the abundance of infiltrating neutrophils. (d) Pearson correlations of CSF3R expression with the abundance of infiltrating neutrophils. (e) Kaplan-Meier survival plot of two tumour groups classified by the mRNA expression of ALOX5AP. F. Kaplan-Meier survival plot of two tumour groups classified by the mRNA expression of CSF3R.